AOPS Precalculus Solutions

PHILLIPS ACADEMY
Solutions Manual
Naoki Sato
# ###^
# K_An.no 1778 • #
# #
# #
# PHILLIPS • ACADEMY #
#
# ## #
###
# OLIVER • WENDELL • HOLMES

# #
# LIBRARY I
#
## # # #
anual
Richard Rusczyk
J
Art of Problem Solving
.
© 2009, AoPS Incorporated. All Rights Reserved.
Reproduction of any portion of this book without the written permission of AoPS Incorporated is strictly prohib¬
ited, except for "fair use" or other noncommercial uses as defined in Sections 107 and 108 of the U.S. Copyright
Act.
Published by: AoPS Incorporated

PO. Box 2185
Alpine, CA 91903-2185
(619) 659-1612
books@artofproblemsolving.com
ISBN #: 978-1-934124-17-8
Visit the Art of Problem Solving website at http: //www. artofproblemsolving. com
Cover image designed by Vanessa Rusczyk using KaleidoTile software.
Printed in the United States of America.
First Printing 2009.

FOREWORD
Foreword
This book contains the full solution to every Exercise, Review Problem, and Challenge Problem in the text
Precalculus.
In most problems, the final answer is contained in a box, like this However, we strongly recommend against
just looking up the final answer and moving on to the next problem. Instead, even if you got the right answer,
read the solution in this book. It might show you a different way of solving the problem that you might not have
thought of.
If you don't understand a solution, or you think you have a better way of solving the problem, or (gasp!) find
an error in one of our solutions, we invite you to come to our message board at
www.artofproblemsolving.com
and discuss it. Our message board is free to use and includes thousands of the world's most eager mathematical
problem-solvers.
CONTENTS
Contents
• • •
Foreword in
1 Functions Review 1
CM
Introduction to Trigonometric Functions 14

CO
Trigonometric Identities 34
4 Applications to Geometry 61
ir>
Parameterization and Trigonometric Coordinate Systems 86
6 Basics of Complex Numbers 104
7 Trigonometry and Complex Numbers 123
IV
CONTENTS
Geometry of Complex Numbers 151
Vectors in Two Dimensions 177
1 Matrices in Two Dimensions 187
1 1 Vectors and Matrices in Three Dimensions, Part 1 207
19
1 mm Vectors and Matrices in Three Dimensions, Part 2 221
IQ
1 w Vector Geometry 241
_ —
'
CHAPTER 1 1
l Functions Review
Exercises for Section 1.1
1.1.1
(a) /(2) = 22 - 2 - 6 = 1-4 |.
(b) Since f(x) = x2 - x - 6 = (x - 3)(x + 2), we have /(x) = 0 when x = 3 and x - -2
(c) f(x - 1) = (x - l)2 - (x - 1) - 6 = x2 - 2x + 1 - x + 1 - 6 = 3x - 4

(d) If /(x) = 6, then x2 - x - 6 = 6, so x2 - x - 12 = 0. Factoring gives (x - 4)(x + 3) = 0, which means x = 4 or
x = -3. Hence, /(4) = /(-3) = 6, so yes , we can have /(x) = 6.
1 2 25
(e) Completing the square, we get /(x) = x-x-6=(x--J - —. Hence, the range of / is [-25/4, +oo)
1.1.2
(a) We can evaluate the absolute value of 1 - x for any real number x, so the domain of / is R . The absolute
value of a real number is always nonnegative, and 1 - x can equal any nonnegative real number, so the
range of / is [0, +oo) .
(b) We can only take the s quare root of a nonnegative real number, so 2 - t must be nonnegative. Hence, the
domain of / i s (-oo,2 . Also, the square root of a nonnegative real number is always nonnegative, so the
range of / is [0, +oo) .
(c) We can evaluate 1 - x2 for any real number x, so the domain of h is R |. Since x2 is always nonnegative,
1 - x2 is always less than or equal to 1, so the range of h is (-oo, 1] .
(d) First, we note that we cannot have u = 0, since this makes the denominator of 1/u equal to 0. We also cannot
have the denominator of the whole function, 1 + 1/m, equal to 0. If 1 + 1/u = 0, then u = —1, so the domain
of g is all real numbers except 0 and -1
To find the range of g, let v = g(u) = Taking the reciprocal of both sides and rearranging gives
11 1 -
1 + - = - u =
M V U V V 1 - V
Hence, for every value of v, there is a corresponding value of u, except when v = 1. Furthermore, as
determined above, u cannot be equal to 0 or -1. There is no real number v that corresponds to u - -1, but
1
CHAPTER 1. FUNCTIONS REVIEW
the real number v = 0 corresponds to u = 0, so v cannot be equal to 0 either. Therefore, the range of g is
all real numbers except 0 and 1
1.1.3 The function (/ • g)(x) = V4 - x • y/2x - 6 is defined only where both the functions f(x) = V4 - x and
g(x) = V2x-6 are defined. The function V4 — x is defined only for x < 4, and the function V2x — 6 is defined
[3,4]
As with / • g, the function
JIM = V4-- X
V2x-6
is defined only where both the functions f(x) = V4 - x and g(x) = V2x - 6 are defined. However, we must also
exclude points where the denominator is 0. In this case, the denominator is 0 when x = 3. Hence, the domain of
the function f/g is (3,4]
The domains of f ■ g and f/g are different because we must exclude those values where g(x) = 0 from the
domain of f/g, but we include these values in the domain of / • g if they are also in the domain of /.
1.1.4 The function

2x - 5
m=
x-8
is defined only where (2x - 5)/(x - 8) > 0. For this inequality to hold, either both 2x - 5 and x - 8 must be positive,
both of them must be negative, or 2x - 5 = 0.
If both 2x - 5 and x - 8 are positive, then x > 5/2 and x > 8, so in this case, we get x > 8. If both 2x - 5 and x - 8
are negative, then x < 5/2 and x < 8, so in this case, we get x < 5/2. Finally, if 2x - 5 = 0, then x = 5/2. Therefore,
the set of x such that (2x - 5)/(x - 8) > 0 is (—oo, 5/2] U (8, +oo), which is the domain of /.
On the other hand, the function

V2x^5
g(x) =
Vx - 8
is defined only where both the functions V2x - 5 and Vx - 8 are defined, and the denominator is nonzero. The
function V2x - 5 is defined for x > 5/2, and the function Vx - 8 is defined for x > 8. The denominator is zero
when x = 8, so the domain of g is (8, +oo).
Thus, the functions / and g have different domains because those values of x for which 2x - 5 and x - 8 are
negative are in the domain of /, but not in the domain o g-
x+1 —X +1 X + 1 x-1
1.1.5 /(x) • /(-x) = - =yj,for |x| * 1.
x-1 -x - 1 1 X +
1.1.6 We break up the interval [-3,4] into the intervals [-3, -2] and [-2,4]. If -3 < x < -2, then -1 < x + 2 < 0,
so 0 < (x + 2)2 < 1. If -2 < x < 4, then 0 < x + 2 < 6, so 0 < (x + 2)2 < 36. Hence, every value in the range of g is in
the interval [0,36].
Intuitively, it seems clear that every value in this interval is in the range. To prove it we let y = (x + 2)2 and solve
for x. Taking the square root of both sides gives x + 2 = ± sfy, so x = ± yjy - 2. Consider the solution x = yjy - 2.
For any y such that 0 < y < 36, we have 0 < yfy ^ 6, so -2 < Vy - 2 < 4. Therefore, the number yjy - 2 is in the
domain of g if 0 < y < 36. So, if we let x = y/y - 2 for any y such that 0 < y < 36, we have /(x) = y. Therefore,
every value in [0,36] is in the range, and our range is [0,36]
1.1.7
(a) We have T( 2,3, -5) = 3 • 23 - (-5) = 3-8 + 5 = [~29
2
Section 1.2
(b) Since T(x, 2,6) - 3x2 — 6, we have 3x2 — 6 = 21. Isolating x2 gives x2 = 9, from which we find x = ±3
1.2.1
(a) If 3x - 7 = 0, then x = 7/3, so the ^-intercept is (7/3,0) I. Since /(0) = -7, the
y-intercept is (0,-7)
f(x) = 3x - 7
(b) If 2 - \x\ - 0, then |x| = 2, so x = ±2. Hence, the x-intercepts are (2,0) and (-2,0) y
Since /(0) = 2, the y-intercept is (0,2)
/(x) = 2 - |x|
(c) Note that /(x) = x2 - 5x + 6 factors as (x - 2)(x - 3). If (x - 2)(x - 3) = 0, then y

x = 2 or x = 3, so the x-intercepts are (2,0) and (3,0) . Since /(0) = 6, the ~r.
y-intercept is (0,6) j
t
r—
\ //
.1.L.
_l.
.
.._..L , J j
1.2.2 /(x) = x2 — 5x + 6
(a) Every vertical line intersects the graph at most once, so the graph represents a function.
(b) Every vertical line intersects the graph at most once, so the graph represents a function. Note that every
vertical line does not need to intersect the graph, such as the line x = 0.
(c) There are many vertical lines that intersect the graph more than once, such as the line x = 1, so the graph
does not represent a function.
1.2.3
(a) All we know about the function / is that /(9) = 2. Hence, if we set x - 3 = 9, then we have x = 12, and
y = /(12 - 3) + 5 = /(9) + 5 = 7. Therefore, a point on the graph of y = /(x - 3) + 5 is (12,7) . We can also
find this point by noting that the graph of y = /(x - 3) + 5 results from shifting the graph of y = /(x) to the
3
right by 3 and up by 5. So, because (9,2) is on the graph of y = /(x), the point (9 + 3,2 + 5) = (12,7) is on the
graph of y = /(x - 3) + 5.
(b) If we set x/4 = 9, then x = 36, and y = 2/(9) = 4. Therefore, a point on the graph of y = 2/(x/4) is (36,4)
We can also see this by noting that the graph of y = 2/(x/4) results from scaling the graph of y = /(x
vertically by a factor of 2 and horizontally by a factor of 4. So, because (9,2) is on the graph of y = /(x), the
point (9 • 4,2 • 2) = (36,4) on the graph of y = 2/(x/4).
(c) Ifweset3x-1 = 9, thenx = 10/3,and y = 2/(9)+ 7 = 11. Therefore, a point on the graph of y = 2/(3x-l) + 7
is (10/3,11)
1.2.4 The graph of y = /(2x) is the same as the graph of y = /(x) compressed by a factor of 2 horizontally, and
the graph of y = /(x - 6) is the same as the graph of y = /(x) shifted 6 units to the right.
Knowing these results, we may think that the graph of y = /(2x - 6) results from compressing the graph of
y = /(x) by a factor of 2 horizontally to give the graph of y = /(2x), and then shifting the result 6 units to the
right. However, this is not true. The first step is correct; compressing the graph of y = /(x) by a factor of two
horizontally gives the graph of y = /(2x). To see how we produce the graph of y = /(2x - 6), we let /z(x) = /(2x).
Then, the graph of y = h(x) is the same as the graph of y = /(2x). To get the graph of y = /(2x - 6), we note that
h(x - 3) = /(2(x - 3)) = /(2x - 6). The graph of y = h(x - 3), which is the same as the graph of y = /(2x - 6), is the
result of shifting the graph of y = h(x), which is the same as the graph of y = /(2x), by 3 units to the right.
Therefore, the answer is (c). The graph of y = /(2x - 6) is the same as the graph of y = /(2x) shifted 3 units to
the right.
1.3.1 If g(x) = 3x + 7, then g(g(x)) = g(3x + 7) = 3(3x + 7) + 7 = 9x + 28 |. If g(x) = -3x - 14, then g(g(x)) =
g(—3x - 14) = —3(-3x - 14) - 14 = 9x + 28 . Hence, both solutions work.
1.3.2
(a) First, we note that the domain of g is (-oo, 1], Since the domain of / is all reals, all outputs of g are in the
domain of /. For x < 1, we therefore have f(g(x)) = /(Vl - x) = 1 - x - 2 Vl - x, where x < 1

(b) The domain of f(g(x)) is the set of values of x such that both g(x) is defined, and g(x) lies in the domain of
/(x). From g(x) = Vl -x, we see that g(x) is defined only for x < 1. The domain of /(x) consists of all real
numbers, so the domain of f(g{x)) is (-oo, 1]
(c) First, /(-2) = (-2)2 - 2 • (-2) = 8. Since g(x) is only defined for x < 1, we see that g(f(-2)) is I not defined
1.3.3 We show that /3(x) is the same as /2(/(x)) by expanding both expressions. By definition, /3(x) = /(/(/(x))).
Since /2(x) = /(/(x)), we may substitute to get /2(/(x)) = /(/(/(x))), so /2(/(x)) = /3(x). Similarly, /(/2(x)) =
/(/(/(x))), so /3(x) and /(/2(x)) are the same, as well.
1.3.4 To get a feel for the problem, we compute /"(x) for the first few values of n:
f\x) = f(x) = ax,

f2(x) = /(/(*)) = f(ax) = a2x,
f3(x) = f(f2(*)) = f(a2x) = a3x,
f4(x) = /(/3(x)) = f(a3x) = a4x,
and so on.
4
Section 1.4
Applying the function / to x multiplies it by a. Therefore, applying the function / to x exactly n times multiplies
it by a exactly n times; in other words, it multiplies it by an. Hence, f n(x) = anx
1.3.5 All the statement "(3,0) is the only ^-intercept of the graph of g" tells us is that g(3) = 0. So, we have
^(3) = /(g(3)) = /(0). This tells us that (3, /(0)) is on the graph of y = h(x), but this doesn't tell us anything about
the intercepts of h. We cannot conclude anything about the intercepts of h from the given information.
1.4.1 Let y — / 1(3). So, we have /(y) = f(f 1(3)) = 3. This means we have
^=3 * y - 1 = 3(y - 2) = 3y - 6 => y

y-2
1.4.2
(a) Let x = /(y) = 2y - 7. Then y = (x + 7)/2, so / x(x) = me + 7)/2
1 __
(b) Let x = f{y) = 2^3. Then 2y + 3 = so y = (J - 3)/2 = (1 - 3x)/(2x). Therefore, /-1 (x) = -~
(c) Let x = /(y) = V2 - y. Then x2 = 2 - y, so y = 2 - x2. Therefore, /_1(x) = 2 - x2. Note that since the range of
f{x) is the set of all nonnegative real numbers, the domain of /-1 is also all nonnegative real numbers. So,
we have / 1(x) = 2 - x2, where the domain of / 1 is [0, +00)
(d) Since /(4) = |5 — 4| = 1 and /(6) = |5 — 6[_ *= 1, there are two different values of x that give the same value of
f{x). Therefore, the function fix) does not have an inverse .
(e) Since /(2) = V4 - 2 + V2 - 2 = x/l and /(4) = V4 — 4 + V4 - 2 = V2, there are two different values of x
that give the same value of fix). Therefore, the function fix) does not have an inverse
(f) Completing the square gives fix) = (x - 3)2 - 6. We might at first think that this function does not have an
inverse, since /(l) = /(5). However, fix) is only defined for x > 4. By considering the graph of fix), or by
noting that the function fix) strictly increases as x increases for x > 4, we see that no two values of x in the
domain of / give the same output from /. In particular, /(4) = -5, so the range of / is [-5, +00). So, / does
have an inverse. But what is it?
To find the inverse of /, we solve x = /(y) = (y - 3)2 - 6 for y in terms of x. Adding 6 to both sides gives
(y - 3)2 = x + 6. Since x is in the range of /, we have x > -5, so x + 6 > 1. Therefore, we take the positive
square root to find y = Vx + 6 + 3. Hence, / !(x) = Vx + 6 + 3, where the domain of / 1 is [-5, +00) . Note
that / *(x) is defined only for x > -5, because this is the range of /.
1.4.3 To find /-1(x), let x = /(y) = ay + b, so y = (x - b)/a. Then the equation fix) - /-1(x) becomes
x-b x b
ax+ b -
a a
This holds for all x if and only if a = 1/a and b = -b/a. From the first equation, we have az = 1, so a = 1 or a = -1.
If a = 1, then the second equation becomes & = -b, so b = 0. If a = -1, then the second equation becomes
b = b, which is true for any real number b. Therefore, the ordered pairs (a,&) that satisfy the problem are
(1,0) and (-1, b) for any any real number b
5
We can check these solutions quickly. For (a, b) - (1,0), we have f(x) = x, so /(/(x)) = f(x) - x, which means
/'_1 (x) = f(x). If (a, b) = (-1, b), then f(x) = -x + b, so /(/(x)) = f(-x + b) = -(-x + b) + b = x, and again we have
f~\x) = f(x).
1.4.4 First, we find y_1(x)- Let x = g(y) = (ay + b)/(cy + d). Then
x(cy + d) = ay + b => cxy + dx = ay + b => cxy - ay = b - dx => y(cx - a) = b — dx

b - dx
—^ y ■
^ cx - a
b - dx
Therefore, we have g 2(x) = -——. We see that g !(x) is defined for all x except when cx - a = 0, or x — a/c
(Note that we must have ad ± be because if ad = be, then g(x) is a constant for all x such that cx + d 0, which
means that g does not have an inverse.)
Review Problems
1.23
(a) Since A(x) = 4x2 + 1 is defined for all real numbers, the domain of A is |R . Since 4x2 can equal any
nonnegative real number, but no negative number, the range of A(x) = 4x2 + 1 is [1, +oo)
(b) The function o(x) = 3 + y/l6 - (x - 3)2 is defined only when (x - 3)2 < 16, so x - 3 must be between -4 and
4. In other words, -4 < x - 3 < 4, so -1 < x < 7. Therefore, the domain of o is [-1,7]
As x varies from -1 to 3, y/l6 - (x - 3)2 varies from 0 to 4, so o(x) = 3 + y/l6 - (x - 3)2 varies from 3 to
7. As x varies from 3 to 7, \Jl6 - (x - 3)2 varies from 4 to 0, so o(x) = 3 + -^16 - (x - 3)2 varies from 7 to 3.
Therefore, the range of o is [3,7]
(c) The function P(x) = 1/(3 + Vx + 1) is defined if and only if x + 1 > 0, or x > —1. Therefore, the domain of P
is [—1, +oo)
Since Vx + 1 can take on any nonnegative value, we see that P(x) = 1/(3 + Vx + 1) varies from 1/3 to 0
(without ever reaching 0). Therefore, the range of P is (0,1/3]
(d) The function S(x) = (12x - 9)/(6 - 9x) = (4x - 3)/(2 - 3x) is defined for all x except when the denominator
2 - 3x is zero. We have 2 - 3x = 0 when x = 2/3, so the domain of S is all reals except 2/3
To find the range of S, we let y = S(x) - (4x - 3)/(2 - 3x). Then we have
2y + 3
y(2 - 3x) = 4x - 3 => 2y — 3xy = 4x - 3 => 3xy + 4x = 2y + 3
~ 3y + 4 ’
Hence, for every value of y except y = -4/3, there is a corresponding value of x such that S(x) = y. Therefore,
the range of S is all reals except -4/3
1.24
(a) The function is defined only when 2x - 5 > 0, or x > 5/2, and the function V9 - 3x is defined only
V2x-5
when 9 - 3x > 0, or x < 3. Combining these, the domain of / is (5/2,3]
(b) The function | yjx - 2| is defined only when x > 0 and the function | Vx - 2| is defined only when x - 2 > 0,
or x > 2. Therefore, the domain of / is [2, +<x>)
6
Review Problems
(c) The function VM ~ 2 is defined only when |x| > 2. This is equivalent to x < — 2 or x > 2. The function V|x - 3|
is defined for all x, since \x - 3| is always nonnegative. Therefore, the domain of g is (-oo, -2] U [2, +oo)
1.25 f{g{x)) = f ( 21
Vx +
1
4)
4 • ggi
+2
_ 8x
2x + 2(x + 4)
8x
4x + 8
2x
x+2
where x £ 4, because -4 is not in the
domain of g.
1.26 The equation /(1) - g(l) + 2 gives us a + b + c = a- b + c +2, so b = 1. Then the equation /(2) = 2 gives
4a + 2fr + c = 2, so 4a + c = 2 - 2b = 0. Therefore, g(2) = 4a - 2b + c = (4a + c) - 2b = 0 - 2 = -2
1.27 For part (a), we shift the graph 3 units to the right. For part (b), we scale the graph vertically away from the
x-axis by a factor of 2, then shift the graph 1 unit upwards. For part (c), we reflect the graph over the y-axis.
For part (d), we reflect the graph over the y-axis, then shift the graph 2 units to the right. To see why we
shift to the right rather than the left, let h(x) = /(—x). Then, the graph of y = /(2 - x) is the same as the graph of
y = h(x - 2). The graph of y = h(x - 2) is the result of shifting the graph of y = h(x) to the right 2 units. So, the
graph of y = /(2 - x) is the result of shifting the graph of y = /(-x) to the right 2 units.
For part (e), we compress the graph horizontally by a factor of 1/2, then shift the graph 1/2 unit to the right.
As with part (d), we can see that this shift is 1/2 unit by letting h(x) = /(2x); therefore, h(x — \) = /(2x - 1), so we
must shift the graph of h(x), which is the same as the graph of y = /(2x), to the right 1/2 unit to get the graph of
y -- /(2x - 1). We then compress the graph of y = /(2x - 1) vertically by a factor of 1/2, then shift the resulting
graph 3 units upwards.
The results for all five parts are shown below.
y = 2/(x) + 1 y = /(2 - x)
1.28 The functions / and g need not be the same. For example, let f(x) - |x| and g(x) = x. Then f(f(x)) = |/(x)| =
||x|| = |x|, and g(f(x)) = g(|x|) = |x|. So, we have f(f(x)) = g(f(x)), but f(x) * g(x).
7
1.29
y
(a) Every vertical line intersects the graph at most once, so the graph can represent a
function. We also see that every horizontal line intersects the graph at most once, —
so the function has an inverse, the graph of which is shown at right. (It is simply —-
the graph in thp prnblpm rpflprtpd ovpr thp linp y — r )
-► v
(b) There are vertical lines that mtersect the graph more than once, so the graph -- .fC
cannot represent a function. / :4: _
(c) Every vertical line intersects the graph at most once, so the graph can represent a -f..
function. However, there are horizontal lines that intersect the graph more than
once, so the function does not have an inverse.
cries-
_
Oil
1.30 From f(a, b, c) - 1, we have = 1, so a - c = b - c, which means a = b. Then f(a, c, b) 0 Jote
that we know that b ± c because f(a, b, c) is defined, so we know that we don't have divisicm by XI (jva lua ting
f(a,c,b).)
1.31 If /(x) = -x2 + bx + c and g(x) = dx + e, then
f(g(x)) = f{dx + e) = -(dx + e)2 + b(dx + e) + c = -d2x2 - 2dex - e2 + bdx + be+ c = -d2x2 + (bd - 2de)x - (e2 - be - c).
For f(g(x)) to be equal to x2 for all x, the coefficient of x2 must be 1. However, there is no real number d for
which -d2 = 1, so it is not possible for f(g(x)) to be x2.
1.32 To find the inverse of f(x) - ax + b, set x - f(y) - ay + b. Then y = ^ | - -a, so we have / x(x) - \x—\.
This is a linear function. We conclude that the inverse of a linear function is always a linear function.
1.33 Consider a point (x,\f(x)\) on the graph of y - |/(x)|. If f(x) > 0, then \f(x)\ = f(x). On the other hand, if
f(x) < 0, then \f(x)\ = -f(x), and the point (x, |/(x)|) = (x, —f(x)) is the reflection of the point (x,/(x)) over the x-axis.
Thus, the graph of y = \f(x)\ can be constructed from the graph of y = f(x) by reflecting over the x-axis the
portion of the graph of y = f(x) that lies below the x-axis. An example is shown below. The dashed lines in the
diagram on the right indicates portions of the graph of y - f(x) that are reflected over the x-axis to produce the
graph of y = |/(x)|.
V = fix) V = \f(x)\
1.34 If y =/(x) = x/(l - x), then
y(l-x) = x =» y - xy - x =» xy + x = y => x(y + 1) = y => x= y

y + i'
Challenge Problems
We compare this to each of the given possibilities:
Vy _
1-1 /y y-i'
_y_
-/(y)
i-y'
-y = _y_
-f(-y) l+y y +1'
-y =_y_
/(-y) l+y y + 1'
y
f(y)
i-y'
Therefore, we have x = -/(-y)

y+i
c2x c2x
1.35 We have /(/(x)) =f(C*/(2* + 3) =
J J ” J \2x + 3/ 2 • cx/(2x + 3) + 3 2cx + 3(2x + 3) (2c + 6)x + 9 ’
If /(/(x)) = x for all x except -3/2, then
—-—-- = x => c2x = x[(2c + 6)x + 9] => c2x = (2c + 6)x2 + 9x.
(2c + 6)x + 9 v ’
This equation holds for all x except -3/2 if and only if 2c + 6 = 0 and c2 = 9. The only value of c that satisfies
both equations is c = | —3 |.
1.36 Let g(x) - f(2x). As discussed in the text, we find the inverse of g(x) by solving the equation x = g(y) for
y in terms of x. From x - g(y), we have x = f(2y). Because / is invertible, x = f(2y) tells us that 2y = /-1(x), so
y - \f~l(x). Thus, the inverse of f(2x) is |/_1(x) , not /-1(2x). We also could have found the inverse using the
fact that the inverse of (f(g(x)) is g l(f l(x)), which we proved in the text. Here, we let g(x) = 2x, so g x(x) = l2x
and the inverse of/(2x) is y-1(/-1(x)) = \ (x), as before.
Note that substituting f(2x) into the function \ (x) gives x: we have \f~1(f(2x)) = l(2x) = x, as expected.
Also, we have / (2 • |/_1(x)) = f(f~l(x)) = x.
1.37 For the functions / and g to be inverses, we also require that g(f(x)) = x for all x in the domain of /. But
g(f(x)) - Vx2 = |x|, which is not equal to x when x is negative, so / and g are not inverse functions.
Since the function f(x) = x2 fails the horizontal line test (for example, /(-1) = /(1) = 1), no function g can be
an inverse of /.
Challenge Problems
1.38 If /(x) = /_1 (x), then /(/(x)) = /(/-1 (x)) = x for all x in the domain of /. We find an expression for f(f(x)) as
follows:
/ 2x + a \ _ 2 ■ (2x + a)/(bx -2) + a _ 2(2x + a) + a(bx - 2)
/(/(*)) “ J {yx _ 2)~ b ■ (2x + a)/(bx -2)-2~ b(2x + a)~ 2 (bx - 2)
4x + 2a + abx -2a _ (ab + 4)x
2bx + ab - 2bx + 4 ab + 4
9
This simplifies to x as long as ab + 4 =£ 0. If ab + 4 = 0, then a = -4/b, and
2x + a _ 2x - 4/b _ 2bx - 4 _ 2(fcx - 2) _ 2

X bx-2 bx-2 b(bx- 2) b(bx- 2) fe'
which is a constant, so / does not have an inverse if ab = -4.
Hence, / x(x) exists and / 1(x) = /(x) for all a and b such that ab £ -4
139
(a) The function f(x + 1) is defined only when -1 < x + 1 < 1, or -2 < x < 0. Therefore, the domain of f(x + 1)
is (-2,0)
(b) The function /(1/x) is defined only when -1 < 1/x < 1. If x is positive, then multiplying all three parts of
the inequality chain by x gives —x < 1 < x. Since x is positive, —x is negative, so the inequality -x < 1 is
always satisfied. Hence, the solution in this case is x > 1.
If x is negative, then multiplying all three parts of the inequality chain by x gives -x > 1 > x. Since x is
negative, the inequality 1 > x is always satisfied. Multiplying the inequality —x > 1 by —1, we get x < —1.
Hence, the solution in this case is x < -1.
Therefore, the domain of /(1/x) is (—oo, -1) u (1, +oo)

(c) The function f(xfx) is defined only when both x > 0 and -1 < y/x < 1. Since \[x is always nonnegative,
the inequality -1 < y/x is always satisfied. For the same reason, we can square both sides of the inequality
y/x < 1 to get x < 1. Therefore, the domain of the function /(y/x) is [0,1)
(d) The function /(fiy) is defined only when — 1 < < 1. We take the cases where x - 1 > 0 and x - 1 < 0
separately.
If x - 1 >0, then multiplying all parts of the inequality -1 < |^ < 1 by x - 1, we get
-x + l<x+l<x-l.
Subtracting x + 1 from all three parts gives —2x < 0 < -2. Therefore, there are no solutions in this case.
If x -1 <0, then multiplying all parts of the inequality -l<jzi<lbyx-l (and changing the directions
of the inequalities because x - 1 is negative), we get
—x +1 >X + 1 > X — 1.
Subtracting x + 1 from all three parts gives -2x > 0 > -2, so x < 0. Therefore, the solution is x < 0 in this
case.
Hence, the domain of /(fzj) is (-oo, 0)
1.40 The function /(x) = V2 - x - x2 is defined only when 2 - x - x2 > 0. Factoring gives (2 + x)(l - x) > 0. Either
both factors must be nonnegative, or both factors must be nonpositive.
If both factors are nonnegative, then 2 + x > 0, so x > -2, and 1 - x > 0, sox<l. Hence, the solution in this
case is -2 < x < 1. If both factors are nonpositive, then 2 + x < 0, so x < -2, and 1 - x < 0, so x > 1. There is no
value of x such that x < -2 and x > 1 simultaneously, so there is no solution in this case. Therefore, the domain of
/(x) is [-2,1]
To find the range of /(x), we complete the square inside the radical:
/(x) = V2 - x - x2 =
10
Challenge Problems
As x varies from 2 to —1/2, the expression 9/4 — (x +1/2)2 varies from 0 to 9/4, so f(x) varies from 0 to 3/2. Then as
x varies from -1/2 to 1, the expression 9/4 — (x + 1/2)2 varies from 9/4 to 0, so f(x) varies from 3/2 to 0. Therefore,
the range of f(x) is [0,3/2]
1.41 Let t = x2 + 1, so x2 = t - 1 and x4 = (t - l)2. Hence,

f(t) = f(x2 + l) = x4+5x2 + 3 = (t-l)2 + 5(t-l) + 3 = t2-2t + l+5t-5 + 3 = t2 + 3t-l.
Substituting t - ,
x2 - 1 we get f(x2 - 1) = (x2 - l)2 + 3(x2 - 1) - 1 c4 - 2x2 + 1 + 3x2 - 3 - 1 = x4 + x2
1.42 Consider a point (x, /(|x|)) on the graph of y = /(|x|). If x > 0, then |x| = x, so (x,/(|x|)) is the same as (x,/(x)).
On the other hand, if x < 0, then |x| = -x, and the point (x, /(|x|)) = (x, /(-x)) is the reflection of the point (—x, /(-x))
over the y-axis.
Thus, the graph of y = /(|x|) consists of the portion of the graph of y = /(x) that is on or to the right of the
y-axis, together with the curve formed when reflecting this portion of the graph of y = /(x) over the y-axis. An
example is shown below.
y y(>>k
7 L 7
\
\\ // A\ /
//
\
7\ !
/ \).
V—
\\/ / \
\i
J\
\
/
/
V 7
*
V = /(*) y = /(M)
1.43 Let x be a root of /(x) = 0. Applying / to both sides, we get f(f(x)) = /(0) = 2003. But f(f(x)) = x for all
values of x, so x = 2003. Therefore, the only root of the equation f(x) = 0 is x 2003
1.44 To find the inverse of /(x), set x = /(y) = (ay + b)/(cy + d). Then
b - dx
x(cy + d) = ay + b => cxy + dx = ay + b => cxy - ay = b - dx cx-a'
b - dx
Thus, the inverse of f(x), if it exists, must be / 1 (x)
cx-a'
To check if this works, we substitute:
a ■ (b - dx)/(cx - a) + b _ a(b - dx) + b(cx - a) (be - ad)x

c- (b - dx)/(cx - a) + d c(b - dx) + d(cx - a) be-ad
This simplifies to x as long as ad - be £ 0. Hence, the function / has an inverse if and only if ad - be ± 0
Note that the function f(x) is well-defined as long as c and d are not simultaneously 0, and the function
b - dx
/-V) =
cx- a'
is well-defined as long as a and c are not simultaneously 0. However, if c = d = 0, then ad-bc = 0, and if a = c = 0,
then ad-bc = 0. Hence, the condition ad-bc ± 0 also ensures that both of these functions are well-defined.
11
1.45 Let (x, 0) be an x-intercept of the graph of /, so f(x) = 0. Then g(x) = /(x)/(|x|) = 0, so (x, 0) is also an
x-intercept of the graph of g. Hence, all x-intercepts of the graph of / must also be x-intercepts of the graph of g.
On the other hand, an x-intercept of the graph of g is not necessarily an x-intercept of the graph of /. For
example, take /(x) = x - 1, so g(x) = /(x)/(|x|) = (x - l)(|x| - 1). Note that g(-l) = 0, but /(-1) = -2^0. Thus, the
point (-1,0) is an x-intercept of the graph of g, but not an x-intercept of the graph of /.
1.46 To construct the graph of y = \f(2x - 2) - 1 from the graph of y = /(x), we describe the construction one
step at a time, working from the inside of the function out.
First, we look at the expression /(2x - 2) = /(2(x - 1)). As described in the solution to Problem 1.2.4, the graph
of y = /(2(x -1)) is the result of scaling the graph of y = f(x) horizontally towards the y-axis by a factor of 1/2 and
shifting the ensuing graph to the right 1 unit. This produces the middle graph below.
Then, the graph of y = |/(2x - 2) is the result of scaling the graph of /(2x - 2) vertically away from the x-axis
by a factor of 3/2, and the graph of y = |/(2x - 2) - 1 results from shifting the graph of y = |/(2x - 2) downward
1 unit. This produces the final graph at right below.
1.47 To find the unique number that is not in the range of /, let y = /(x) = (ax + b)/(cx + d). Then
b - dy
y(cx + d) = ax + b => cxy + dy = ax + b cxy - ax = b - dy x-
cy-a'
Hence, for every value of y, there is a corresponding value of x such that /(x) = y, except when the denominator
cy-a is equal to 0. Therefore, the unique number that is not in the range of / is a/c.
Now we compute /(/(x)):
ax + b\ a - (ax + b)/(cx + d) + b a(ax + b) + b(cx + d)

/(/(*)) = /
cx + dJ c ■ (ax + b)/(cx + d) + d c(ax + b) + d(cx + d)
(a2 + bc)x + (ab + bd)
(ac + cd)x + (be + d2)'
(a1 + bc)x + (ab + bd)

If f(f(x)) = x for all x, then = x, so (a2 + bc)x + (ab + bd) - (ac + cd)x2 + (be + d2)x for all x.
(ac + cd)x + (be + d2)
Then the coefficient of x2 must be 0, so ac + cd - 0, which means c(a + d) - 0. Since c is nonzero, we have
a + d = 0, so d - -a. Furthermore, if d = -a, then
(a2 + bc)x + (ab + bd) (a2 + bc)x + (ab - ab) (a2 + bc)x
X (ac + cd)x + (be + d2) (ac - ac)x + (a2 + be) a2 + be
12
Challenge Problems
ax + b ax + b
as desired. Therefore, f(x) =
cx + d cx — a
Since /(19) = 19, we have
19 a + b
= 19 192c - 19a = 19 a + b 19 2c = 2-19 a + b.
19 c - a
Also, /(97) = 97, so we have
9 7a+ b
= 97 972c - 97a = 97a + b 972c = 2-9 7a+ b.
97c - a
Subtracting the first equation from the second, we get
a 97 + 19
(972 - 192)c = 2(97 - 19)a (97 + 19)(97 - 19)c = 2(97 - 19)a 58
Here is a faster way to derive the same equations. Since /(/(x)) = x for all x, we have /(/(0)) = 0. Since
/(0) = b/d, we have
^
mm=nm
r/ r/nw
= a-b/d + b ab + bd
so /(/(0)) = 0 gives (ab + bd)/(bc + d2) = 0. Therefore, we have ab + bd = 0, so b(a + d) = 0. Since is nonzero, we
have a + d = 0, so d = -a.
Now, we know that x = 19 and j = 97 satisfy the equation /(x) = x. Writing this equation out, we get
ax + b
=x => x(cx + d) = ax + b cx + (d- a)x-b = 0.
cx + d
The sum of the roots of this quadratic is -(d-a)/c = la/c. We know that the roots of this quadratic are 19 and 97,
since these are the solutions to /(x) = x. Therefore, we have 2a/c = 19 + 97 = 116, which gives us a/c = 116/2 = 58.
13
CHAPTER 2. INTRODUCTION TO TRIGONOMETRIC FUNCTIONS
CHAPTER 2 1
l Introduction to Trigonometric Functions
2.1.1
(a) Let P be the terminal point of 120° and let S be the foot of the perpendicular from P to the x-axis. From 30-
60-90 triangle POS, we have PS = V3/2. Since 120° lies in the second quadrant, we have sin 120° = V3/2
Figure 2.1: Diagram for Part (a) Figure 2.2: Diagram for Part (b)
(b) Let P be the terminal point of 60°, and let S be the foot of the perpendicular from P to the x-axis. From
30-60-90 triangle POS, we have PS = V3/2 and OS = 1/2, so
cos 60° OS _ 1/2 V3

cot 60° =
sin 60° PS ~ V3/2 3
1 _
(c) Since cos 45° = V2/2 = 1/ V2, we have sec 45° = cos 45°
V2
1/V2
14
Section 2.1
(d) Let P be the terminal point of 210° and let S be the foot of the altitude from P
to the x-axis. From 30-60-90 triangle POS, we have OS = V3/2 and PS = 1/2.
Since 210° lies in the third quadrant, we have cos 210° = - V3/2 and sin 210° =
-1/2, so
sin210c -1/2 _ 1 V3
tan 210° =
cos210c -V3/2■" V3 3
(e) The terminal point of 180° is (-1,0), so cos 180° = -1
(f)
anuuuc Hum r iu uit x-dxis. men = ou , so us = i/z ana ri>
Since 300° lies in the fourth quadrant, we have sin 300° = — V3/2, so
esc 300°
sin 300° -V3/2 x
2.1.2
(a) The terminal point of -90° is (0, -1), so cos(-90°) - 0

(b) Since 36000° = 100 • 360°, the terminal point of 36000° is the same as the terminal point of 0°. Therefore,
tan 36000° = tan0° 0
(c) Since 750° = 2 • 360° + 30°, the terminal point of 750° is the same as the terminal point of 30°. Therefore,
sin 750° = sin 30° = 1/2
(d) Since -1200° = -4 • 360° + 240°, the terminal point of -1200° is the same as the
terminal point of 240° = 180° + 60°. We let P be this terminal point, and let S be
the foot of the altitude from P to the x-axis. Then ZPOS = 60°, so OS = 1/2 and
PS = V3/2. Since 240° lies in the third quadrant, we have cos 240° = -1/2, so
sec(-1200°) = sec 240° = -2

cos 240° -1/2
2.1.3 The angle 293° lies in the fourth quadrant, so sin 293° is negative. The angle -68° lies in the fourth quadrant,
so cos(-68°) is positive. The angle 206° lies in the third quadrant, so cot 206° = is positive. The angle 90.5°
lies in the second quadrant, so esc 90.5° = sing05<, is positive.
Thus, cos(-68°), cot 206°, and esc 90.5° are positive.
15
2.1.4 If cos 0 = x/l/2, then cos2 0 = 2/4 = 1/2. We know that cos2 0 + sin2 0 = 1, V
so sin2 6 = 1 - 1/2 = 1/2, which means sin0 = ± x/l/2 = + V2/2. Hence,
(cos 0, sin 0) = (V2/2, a/2/2) or (V2/2, - a/5/2).
We know that (V2/2, V2/2) is the terminal point of 45°. However, the point
(V2/2, V2/2) is also the terminal point of 45° + 360° = 405°, 405° + 360° = 765°,
and 765° + 360° = 1125°. The angles in this list that are in the range 0° < 0 < 900°
are 45°, 405°, and 765°.
We know that (V2/2, - V2/2) is the terminal point of -45°. However, the point
(V2/2, - V2/2) is also the terminal point of -45° +360° = 315°, 315° + 360° = 675°, )
and 675° + 360° = 1035°. The angles in the range 0° < 0 < 900° are 315° and 675°.
Therefore, the solutions such that 0° < 0 < 900° are 0 = 45°, 315°, 405°, 675°, and 765°. There are [IF] such
solutions.
2.1.5 The range of the function sin3x is [—1,1]. Then the range of the function 2 sin 3x is [-2,2], and so the range
of the function f(x) = 7 + 2 sin 3x is [5,9]
2.2.1
lln
(a) 330° = 330° • ^ = 6
3n
(b) 270°
u = 270°
a/\j • —
36Qo -
2
(c) -315° = -315° • ^ = Ze

4
IOOti
(d) 2000° = 2000° • =
2.2.2
(a) 371 = 371 • ^ = 540°

2n - 2n 36CT .
(b) 3 3 ' 2n '
120 °
.7n _ _7n . 3601 -

(c) 5 5 2n
-252°
540°
(d) -3 = -3-^ = 71
16
Section 2.2
2.2.3 Converting the angle measures to degrees, we find that:
n 71
30° = 30° •
180° 6 '
71 n
45° = 45° •
180° 4'
u TL
60° = 60°•
180° 3'
n TL
90° = 90°•
180° 2 '
Each leg of the 45-45-90 triangle has length and the legs of the 30-60-90 triangle have length \ and with
the shorter leg opposite the smaller acute angle. The triangles are drawn below.
TL Tl TL TL TL TL .
Figure 2.3: triangle Figure 2.4:
TTi 6T2 trlangle
2.2.4
(a) Converting j to degrees, we find j - j • = 45°. Then sin j = sin45° = V2/2
(b) The terminal point of 7n is the same as the terminal point of 7n - 3(271) = n, which is (-1,0), so sin 71 = 0.
Therefore, esc 7n - esc n, which is undefined .
(c) We let P be the terminal point of 5n/6 and S the foot of the altitude from P to the x-axis. Then ZPOS = 7t/6,
so OS - V3/2 and PS = 1/2. Since 5tl/6 lies in the second quadrant, we have cos(57i/6) = — V3/2 and
sin(57i/6) = 1/2, so
5tl
57T sin 1/2 _ 1 V3
tan
cos ^ V3/2 " vi: 3
Figure 2.5: Diagram for Part (c) Figure 2.6: Diagram for Part (d)
(d) We let P be the terminal point of -7i/3 and S be the foot of the altitude from P to the x-axis. Then ZPOS = n/3,
so OS = 1/2 and PS = V3/2. Since -n/3 lies in the fourth quadrant, we have cos = 1/2
2.2.5 Since tan t = we see that tan t is defined as long as cos t t 0. But cos t = 0 if and only if t is an odd
multiple of f, that is, if t = (2<^1)n for some integer k. So, tan t is defined if and only if t is not of the form (2^1)7T.
17
Hence, /(x) = 3 tan(2x - f) is defined if and only if 2x - 5 is not of the form -2^1)7T. Therefore, the values of x
that we must omit from the domain are the values of x such that 2x - | = ^y-— where k is an integer. Solving
this equation for x gives x = (6k^n.
Therefore, the domain of /(x) = 3 tan(2x - f) is all real numbers except those numbers of the form (<lk^',7T where
k is an integer.
2.2.6
(a) Since 40° lies in the first quadrant, tan 40° is positive.
(b) Jake probably computed tan 40, where the angle 40 is in radians. Indeed, tan 40 = -1.11721 to five decimal
places. Jake should have either set his calculator to degrees mode, or computed tan(40° • Jy) = tan y
instead.
2.3.1
(a) The graph is shown at right.
(b) Since secx - and cos x has period 2n, sec x also has period 2n
(c) Since sec x = and esc x = and the graph of y = cos x is the graph
of y = sin x shifted | to the left, the graph of y -- sec x is the graph of
y ~ esc x shifted j to the left.
2.3.2
(a) The graph is shown at right.

—
(b) Since cotx - and tan x has period n, cotx also has period 71
x v\ —
n
V V r. n
2.3.3 Below, the graph of y = cos x is dashed and the graph of y = sin x is solid.
It appears that the two graphs intersect twice in each period. To see why this is the case, consider the unit
circle. If cos 6 = sin 0, then the coordinates of the terminal point of the angle 6 are the same. All the points whose
18
Section 2.4
coordinates are the same lie on the graph of x = y. The intersection of this line with the unit circle gives us the
two terminal points whose coordinates are the same, (^/ x) and
We know that (-y, -^) is the terminal point of j. However, (4jp, -^) is also the terminal point of j + Ink for
any integer k. The smallest number of this form in the interval [0,100] is |, and the largest number of this form is
f + 271 • 15 ~ 95.03. Hence, we can take k = 0,1,, 15, for a total of 16 different values of k.
We know that is the terminal point of However, is also the terminal point of
if + 27ik for any integer k. The smallest number of this form in the interval [0,100] is , and the largest number
of this form is ^ + 271 • 15 ~ 98.17. Hence, we can take k = 0,1,..., 15, for a total of 16 different values of k.
Therefore, the graphs of y - sin x and y = cos x intersect 16 + 16 = 32 times in the interval [0,100].
2.4.1 The range of sinx is [-1,1]. If a is nonnegative, then the range of /(x) = a sinx is [-a,a], so the amplitude
of /(x) is [a - (-fl)]/2 = a = \a\. If a is negative, then the range of /(x) = a sinx is [a, -a], so the amplitude of /(x) is
(-a - a)/2 - -a - \a\. In either case, the amplitude of /(x) is \a\.
2.4.2 The graph of y = 3+sin x is obtained by shifting

the graph of y = sinx up by 3 units. This graph is
shown at right.
/•
in n 7T 271
2.4.3
(a) The graph of y = cos 2x is obtained by compressing the
graph of y = cos x horizontally towards the x-axis by a
factor of 2. This graph is shown at right.
(b) The graph of y = -3 cos 2x is obtained by stretching the

graph of y = cos 2x vertically by a factor of 3, and then
reflecting the result over the x-axis. This graph is shown
at right.
(c) We note that -3 cos 2x is equal to -3 when x = 0. Cor¬

respondingly, the function -3cos(2x - f) is equal to -3
when 2x - f = 0, or x = f. Therefore, \ is the phase
shift of the graph y = -3cos(2x - f), and we find this
graph by shifting the graph of y = -3 cos 2x to the right
by f. The resulting graph of y = -3 cos(2x - f) is shown
at right.
19
2.4.4 The amplitude of the function corresponding to the given graph is 3. The amplitude of the function
g(x) = f sin ux is f when f is positive, so we have t - 3.
If t = 3, then g(x) = 3 sin ux, so the graph of g(x) is obtained by stretching the graph of y - sinx vertically by
a factor of 3 and compressing it horizontally by a factor of u. To find u, note that the graph completes 5 cycles in
the interval [0,2n]. Therefore, u = 5, and g(x) = 3 sin 5x, so (f, u) = (3,5)
2.4.5 We start with the function cosx, which has period 2n and amplitude 1. The graph of cos 2x is obtained
by compressing the graph of cos x horizontally by a factor of 2, so the function cos 2x has period ^ = n and
amplitude 1.
The graph of -3 cos 2x is obtained by stretching the graph of cos 2x vertically by a factor 3, and then reflecting
it in the x-axis. The period is still tl, but the amplitude is now 3.
The function -3 cos 2x is equal to -3 when x = 0. Correspondingly, the function -3 cos(2x + n) is equal to -3
when 2x + n = 0, or x = — The function -3cos(2x + n) also equals -3 when 2x + n = 2tl, or x = . So, we see f
that we can view the graph of y = -3cos(2x + n) as either a rightward or leftward shift of y = -3cos(2x) by |.
Therefore, we can say that the phase shift of -3cos(2x + n) is either —j or |. (Either is acceptable.) The period
and amplitude are still n and 3, respectively.
Finally, adding 4 does not change the period, amplitude, or phase shift. Therefore, the function f(x) =
-3 cos(2x + n) + 4 has period tl, amplitude 3, and phase shift | or -j
2.4.6 Because the distance between the block and the ceiling varies sinusoidally with time, the distance is
f(t) = a sin(bt + c) + d, where a, fa, c, and d are constants. The middle point of the block's path is 1.5 meters from the
ceiling, and it varies 0.3 meters above and below this point. So, the amplitude is 0.3, and /(f) must equal 1.5 when
a sin (bt + c) is 0 (when the block is at the midpoint of its path). Therefore, the function is /(f) = 0.3 sin(faf + c) + 1.5
for some values of b and c.
Suppose we let f = 0 at a point when the block is closest to the ceiling, so we must have /(0) = 1.2. Therefore,
we must have sin(faf + c) = -1 when f = 0, which means we can take c = — |, and we have
/(f) = 0.3 sin (bt - ~ j + 1.5.
Finally, we turn to the last piece of information we have: that the block goes from the high point to the low point in
1 second. This means that it goes from high to low and back to high again in 2 seconds, so its period is 2 seconds.
Since the period of /(f) is when b is positive, we must have b - tl, and we have
/(f) = 0.3 sin (nt _ + 1-5.
So, 2.7 seconds after the block is at a high point, the distance between the block and the ceiling is
/(2.7) = 0.3 sin (2.7n - + 1.5 * 11.68 feet
2.4.7
(a) The graph completes one cycle in the interval [0, 3tt], so the function has period 371. But the function cos ax
has period so - 3tl, which means \a\ = 2/3. Since a > 0, we have a = 2/3
(b) The function sin(£>x + c) has the same period as the function sin fax, which has period . Since the period
must be 3n and fa > 0, we have fa = 2/3 .
(c) If f{x) = sin(fax + c), then we have /(0) = sin c. Since the graph passes through (0,1), we must have sin c - 1.
There are no other restrictions on c, so there are infinitely many possible values of c. Since the only restriction
Tl
on c is that sin c = 1, the permissible values of c are the numbers of the form — + 2kn, where fa is an integer
20
Section 2.5
2.4.8 We claim that the function /(x) = 2 sin x - tan x has period 2n.
First, sinx has period 2rr and tanx has period n, so sin(x + 2n) = sinx and tan(x + 2n) = tan(x + n) = tanx,
which means
/(x + 27i) = 2 sin(x + 2n) - tan(x + 2n) = 2 sin x - tan x = /(x).
This tells us that / is periodic, and that the period of / is no greater than 2n.
Let d be the period of /(x) = 2 sin x - tan x, so /(x + d) = /(x). To show that the period of / is 2n, we must show
that d cannot be less than 2n. Note that
„ . sinx 2cosxsinx - sinx

f(x) = 2 sin x - tan x = 2 sin x - —— =-
cos x cos x
so f{x) is undefined whenever cos x = 0. But if /(x) is undefined for some value of x, then /(x + d) must also be
undefined. (Otherwise, the equality /(x + d) = /(x) could not hold.) In other words, if cos x = 0, then cos(x + d) = 0.
The values of x such that cosx = 0 are the integer multiples of n. So, d must be an integer multiple of n. We have
already seen that f(x + 2n) = /(x) for all x, so the only other possibility for the period is tl. To show the period is
not 7i, we can note that /(f) = 2sin f - tan f = V2 - 1 and / (^) - 2sin - tan ^ - —V2 - 1. Therefore, we
have / (f) ^ / (f + 7i), which means the period is not tl.
We conclude that the period of fix) = 2 sin x - tan x is 2n
2.5.1
(a) Since sin f = 0.5 and -f < f < §, we have arcsin0.5 =
/ V2) - 3n
(b) Since cos ^ and 0 < ^ < 7i, we have arccos (-^r)
V
=
2 ) ~ 4
(c) Since tan (-f) = -1 and -f < -f < f, we have arctan(-l) =

(d) Since cos | = 0 and 0 < f < 7i, we have arccos 0 =
(e) Since cot f = ^ and 0 < f < 7i, we have arccot /p =
(f) We seek an angle 6 such that esc 6 = -2. Since esc 6 = we have ^ = -2, so sin0 Since
sin(-f) = and -f 6 [-f, 0) U (0, f ] (the range of esc-1 x), we have csc"1(-2) =
2.5.2 Let 6 = arcsin0.3, so -f < 6 < f and sin 0 = 0.3 = 3/10. Then
91
cos2 0 = 1- sin2 0 = 1-
10/ 100 100
V9l
Since - f < 0 < f, we have cos 0 > 0, so cos(arcsin 0.3) = cos 0 = i/ ^ = 10
2.5.3 Let 0 = arctan(tan ^f), so -f < 0 < f and tan 0 = tan Since tan 0 has period n, and
63n ixtt 2ti

T = 13n"5-'
271
we conclude that tan ^ = tan (-f). Also, -f < -f < Therefore, 0 = ‘ 5
21
2.5.4 Since tan ~ is undefined, the range of arctan cannot include |, so in particular, it cannot be the interval
[0,7t). But even if we removed the value \ from this interval to make the range [0, f) U (|, n), we would face
another problem: this choice would introduce a discontinuity (that is, a gap) in the graph of y = arctan x.
Note that tanO = tanrc = 0. For small positive values of x, arctan;*: would be close to 0. However, for small
negative values of x, arctanx would be close to n.
This kind of behavior is unnatural, and can easily be avoided by setting (-|, |) to be the range of arctan x. The
graph of arctan x is shown below.
2.5.5 The range of the arccos function is 0 < 6 < n. But sin 6 > 0 for all 0 < 6 < tl, and sin 6 ui Liy
and 6 = n. Since cos 0 = 1 and cos n = -1, the only values x for which sin(arccos x) < 0 are X = 1 and -1
The range of the arcsin function is -f <0< n

2’
But cos 6 > 0 for all — | < 0 < |, so in particular, there are
no values x for which cos(arcsin x) < 0.
2.5.6 If y = arctan x, then tany = x and < y < ~. Then x = tan y = Squaring, we get x2 -
Solving for sin2 y, we find sin2 y = . Taking the square root of both sides, we get
x
sin y = ± = +
Vx2 + 1'
The term Vx2 + 1 is always positive. If 0 < y < f, then sin y > 0 and x = tan y > 0, so in the equation above, the
sign of sin y matches the sign of x. If -f
< y < 0, then sin y < 0 and x = tan y < 0, so again, the sign of sin y
matches the sign of x. Since siny always matches the sign of x, we can always express siny as the "+" result of
the "±" in our expression above for sin y. So, we have
x
siny =
Vx2 + 1
2.5.7 The solution in the text includes t = ^ while Kira's solution includes t = ^ We will prove
that these two expressions produce the same set of solutions as follows.
Suppose k = a for some integer a in the text's solution, and k - b for some integer in Kira's solution. We wish
to show that for any integer a, there is a value of b such that these two solutions are the same. So, for a given a, we
want to show that there is an integer b such that
2571 i 2an _ n ( 2bn

~36~ + IT ~ 36 + ~3~'
Solving for b, we find b = a + 1. So, for every solution the form in the text produces, Kira's form also produces
that solution. Conversely, we can see that for any integer b that we choose for Kira's form, setting a = b — 1 in the
form in the text produces the same solution. Putting these observations together, we see that the two expressions
produce the same set of solutions. So, Kira's answer is correct.
22
Review Problems
Review Problems
2.39
(a) Let P be the terminal point of 150° = 180° - 30° and let S be the foot of the perpendicular from P to the x-axis.
Then ZPOS = 30°, so OS = V3/2 and PS = 1/2. Since 150° lies in the second quadrant, cos 150° - V3/2
y y
(b) Let P be the terminal point of 30° and let S be the foot of the perpendicular from P to the x-axis. From
30-60-90 triangle POS, we have OS = V3/2 and PS = 1/2, so tan 30° = PS/OS = 1/V3 = V3/3
(c) The terminal point of 90° is (0,1), so sin 90° = 1, which means esc 90° = 1/ sin 90° = |T~|.
(d) The terminal point of 270° is (0, -1), so cos 270° = 0 and sin 270° = -1. Therefore, tan 270° is undefined
(e) Let P be the terminal point of 315° = 360° - 45° and let S be the foot of the perpendicular from P to the
x-axis. Then ZPOS = 45°, so OS = V2/2 and PS = V2/2. Since 315° lies in the fourth quadrant, we have
cos 315° = V2/2
y y
Figure 2.9: Diagram for Part (e) Figure 2.10: Diagram for Part (f)
(f) Let P be the terminal point of 225° = 180° + 45° and let S be the foot of the perpendicular from P to the
x-axis. Then LPOS = 45°, so OS = V2/2 and PS = V2/2. Since 225° lies in the third quadrant, we have
sin 225° - V2/2
23
2.40
(a) Let P be the terminal point of -120° = -180° + 60° and let S be the foot of the
perpendicular from P to the x-axis. Then ZPOS = 60°, so OS = 1/2 and PS =
V3/2. Since -120° lies in the third quadrant, we have sin(-120°) = - V3/2
(b) Since -540° = -2 • 360° + 180°, the terminal point of -540° is the same as the
terminal point of 180°, which is (-1,0). Therefore, cos(-540°) = | -11.
(c) Since 1320° = 3 • 360° + 240°, the terminal point of 1320° is the same as the
terminal point of 240° = 180° + 60°. Let P be this terminal point and let S be the
foot of the perpendicular from P to the x-axis. Then ZPOS = 60°, so OS = 1/2
and PS = V3/2. Since 240° lies in the third quadrant, we have cos 240° = -1/2
and sin 240° = - y/3/2, so
sin 240° -V3/2

tan 1320° = tan 240°
cos 240° - 1/2
(d) Since 660° = 2 • 360° - 60°, the terminal point of 660° is the same as the terminal point of -60°. Let P be this
terminal point and let S be the foot of the perpendicular from P to the x-axis. Then IPOS = 60°, so OS = 1/2
and PS = V3/2. Since -60° lies in the fourth quadrant, we have cos(-60°) = 1/2, so
1
sec 660° = sec(-60°) =
cos(-60°)
2.41 As 0 increases from 0 to n, cos 0 varies from 1 to -1. Furthermore, it takes on each value in the interval
[-1,1] exactly once, including -0.1. As 0 increases from n to 2n, cos 0 varies from -1 to 1, in the same way. In
general, for each integer n, as 0 increases from nn to (n + 1)ti, cos 0 takes on each value in the interval [-1,1]
exactly once.
Thus, we can divide the interval [0,7n] into the seven intervals [0, n\, [n, 2n],..., [671,7n\. On each of these
intervals, cos 0 takes on the value -0.1 exactly once. Therefore, there are [7] values of 0 in the interval 0 < 0 < 7tl
for which cos 0 = -0.1.
We can also tackle this problem by thinking about the unit circle. The angles 0 for which cos 0 = —0.1 are those
angles whose terminal points have x-coordinate equal to -0.1. The intersections of the graph of x = -0.1 with
the unit circle give us the two such points. One is in the second quadrant and one is in the third quadrant. So,
there are two values from 0 to 2n for which cos 0 = -0.1, one between 0 and n and the other between n and 2n.
Since the period of cosine is 2n, we have cos 0 = -0.1 for exactly one value of 0 between kn and (k + l)n for every
integer k. There are clearly 7 such intervals from 0 to 7n, so there are 7 values of 0 in the interval 0 < 0 < 7n for
which cos 0 = -0.1.
24
Review Problems
2.42 The function tan x is defined for all x, except when x is an odd multiple of in other words, when x =
for some integer n. Hence, in order for /(f) = 2 tan(31 - n) to be defined, we cannot have 3f - n = (2n+^n for some
integer n. Solving for f gives t = -2>l^3)n. Therefore, tan(3f - tl) is defined for all f except f = f2)y—, where n is an
integer. So, the domain of /(f) is all real numbers except numbers of the form (2n+2)n where n is an integer.
The range of /(f) is all real numbers, since tangent can output any real number.
2.43
4n
(a) 144° = 144° • ^ = 5
28n
(b) 336° = 336° • ^ = 15
2.44
5n _ 5n 360° _
(a) 2 ~ 2 ‘ 2n ~
450°
\\n _ ll7i 3601 -
(b) 3 “ 3 In
660°
2.45
(a) Let P be the terminal point of y = 7i - | and let S be the foot of the perpen¬
dicular from P to the x-axis. Then ZPOS = ?, so OS = -y and PS = y. Since
3n 3n V2
4 2
(b) 0
The terminal point of -n is (-1,0), so sin(—n) = [0
(c) Let P be the terminal point of ^ = n + f and let S be the foot of the perpen¬
dicular from P to the x-axis. Then ZPOS = f, so OS - y and PS — y . Since
y lies in the third quadrant, we have cos y = -y and sin y = -y, so
571
5n cos 4 -V2/2
cot — = -r—
4 sin y - V2/2
=0 -
(d) Let P be the terminal point of ^ = n + f and let S be the foot of the perpendicular from P to the x-axis.
Then ZPOS = f, so OS = & and PS = Since ^ lies in the third quadrant, we have cos7f = -^ so
7n _2_ 2V3
sec
-V3/2 V3
25
2.46 Since n < 1.3n < y, the angle 1.3rt lies in the third quadrant, so cos 1.3tz is negative. Since -2n < -6 < —f,
the angle -6 radians lies in the first quadrant, so sin(-6) is positive. Since n < y < y, the angle y lies in the third
quadrant, so tan y is positive. Since y < 6 < 2n, the angle 6 radians lies in the fourth quadrant, so sec 6 = 1/ cos 6
is positive.
Therefore, sin(-6), tan y, and sec 6 are positive.
2.47 Because 3 is between \ and n, the angle 3 radians is a second quadrant angle, which means that sin 3 is
positive. Since 4 is between n and y, the angle 4 radians is a third quadrant angle, which means sin 4 is negative.
So, sin 3 + sin 4 is . . . Uh-oh, we need more information! Fortunately, we can get it. Since 3 is between y and n,
we know that 0 < sin3 < y. Since 4 is between y and y, we know that -1 < sin4 < -y. Aha! Since sin3 < y
and sin 4 < —-y, we have sin 3 + sin 4 < 0. Therefore, sin 3 + sin 4 is negative . (Basically, all we're doing here
is noticing that 3 radians is a second quadrant angle, 4 radians is a third quadrant angle, and 4 is farther from n
than 3 is.)
2.48 In order to have sin(sin x) - 1, we must have sinx = | + 2kn for some integer k. The smallest positive value
| + 2kn can have is y However, | is greater than 1, so we cannot have sin x = y Similarly, the greatest negative
value of \ + 2kn occurs when k = -1, which gives us sinx = -y. However, sinx cannot be less than -1, so this
equation also has no solutions. Therefore, sin(sinx) is never equal to 1.
2.49 The angle y in degrees is y • ^ ^ in radians. Therefore, the line of code should be changed to
y = t * sin(y * pi/18Q) .
2.50
_ cosx
(a) Since cot x = tanx sinx
, cot x is undefined whenever sin x = 0. In turn, sin x = 0 if and only if x is a multiple
of 71; in other words, x - nn for some integer n. Therefore, the asymptotes of the graph of y = cotx are the
lines x = nn, where n is an integer
(b) By part (a), cot3x is undefined whenever sin3x = 0. In turn, sin3x = 0 if and only if 3x is a multiple of 7i;
in other words, 3x = nn for some integer n. Solving for x, we find x = y. Therefore, the asymptotes of the
x = y, where n is an integer
2.51 Note that /(0) = a sin 0 = 0 and g(x) = a cos 0 = a, so the graph of /(x) passes through the origin and the
graph of g{x) does not pass through the origin. This gives us a quick way to identify which graph corresponds to
which function.
2.52 We start with the function sinx, which has period 2n and amplitude 1. The graph of sin | is obtained by
stretching the graph of sin x horizontally away from the y-axis by a factor of 4, so the function sin | has period
A • 2n = 8n and amplitude 1.
Tire graph of -2 sin | is obtained by stretching the graph of sin | vertically away from the x-axis by a factor 2,
and then reflecting it over the x-axis. The period is still 8n, but the amplitude is now 2.
Finally, the function -2 sin \ is equal to 0 when x = 0. Correspondingly, the function g(x) = -2 sin is
equal to 0 when f-f = 0, or x = y. Therefore, the phase shift of g(x) = -2sin(| - |) is y . The period and
amplitude are still 8n and 2 , respectively.
2.53 We start with the function tanx, which has period n. The graph of tan3x is obtained by compressing the
graph of tan x horizontally towards the y-axis by a factor of 3, so the function tan 3x has period 7T3 *
The graph of -2 tan 3x is obtained by stretching the graph of tan 3x vertically away from the x-axis by a factor
of 2, and then reflecting it over the x-axis. The period is still y
26
Review Problems
Finally, the graph of the function g(x) - -2 tan(3x - n) is obtained by translating the graph of -2 tan 3x by the
phase shift, which does not affect the period. Therefore, the period of g(x) is
2.54
(a) The graph of y = sin(| + x) is a ~ leftward shift of the graph of y = sin x, the result of which is shown below.
(b) If /(x) - (| +x), then /(-x) = sin(| -x), so the graph of y - sin(| -x) is the graph of y = sin(| +x) reflected
over the y-axis. Reflecting the graph from part (a) over the y-axis gives us
(c) The graph of y = sin(| - x) appears to be exactly the same as the graph of y = cosx, so it appears that
sin(| - x) = cosx for all x. We'll prove this relationship in the next chapter. See if you can do it on your
own!
2.55 We produce the graph of y = 2 sin y by scaling the graph of y = sin x vertically from the x-axis by a factor
of 2, and horizontally away from the y-axis by a factor of 3/2, producing the graph below.
We note that if /(x) = 2 sin y, then /(x) = 0 for x = 0. If g(x) = 2sin (y - rr), then we have g{x) - 0 for x = y.
Therefore, we can shift the graph of y = 2 sin y by y to the right to produce the graph of y = 2 sin (y - n). The
resulting graph is shown below.
27
2.56 The given graph passes through the point (0,-2), so h(0) = —2. But h(x) = acosbx, so h(0) = acosO - a.
Hence, a = -2
The graph completes 3 cycles between -2n and 271, so the period is [In - (-2rt)]/3 = But the period of
h(x) - a cos bx is so ^ Solving for \b\, we find |b\ = 3/2. Therefore, b = 3/2 or -3/2
2.57
2n
(a) Since cos = -0.5 and 0 < < n, arccos(-0.5) 3
(b) Since sin(-|) = -1 and -f - _f - 2' arcsin(-l) -
(c) Since tanO = 0 and -j < 0 < |, tan-1 0 - 0

V3\ _
(d) Since cos f = ^ and 0 < f < n, arccos (^)
2.58 Let 9 = arctan2, so -\ < 9 < \ and tan@ = 2. Then we have 2 = tan0 = Squaring both ends,
_ sin2 6 sin2 6
we get 4 cos2 6 1-sin2 0
. Solving for sin2 9, we find sin2 9 - I. Taking the square root of both sides, we get
smO = ±-| = ±^.
We know tan 9 = 2. If < 9 < | and tan 9 is positive, then 0 < 9 < so sin 9 is also positive. Therefore,
sin 9 =
2.59 If 3 cot(4k - n) = V3, then cot(4k - n) = V3/3, so
tan(4L - n) = = V3.
cot(41c - 7i) a/3
Since tan x has period n, this equation is equivalent to tan 4k = V3.
We know that arctan V3 = f. Since tan x has period n, we conclude that tan v = V3 if and only if
n (3 n + l)n
x = — + nn ----
3 3
for some integer n. Therefore, if tan 4k - V3, then 4k = --T1— for some integer n. This means the solutions are of
(3n+l)n
the form k = 12
, where n is an integer.
2.60 The graph of y ~ arctan x is shown below:
The graph of y = arctan x is the reflection over the line y = x of the portion of the graph of y = tan x on the
interval -| < x < Since the graph of y = tanx has the vertical asymptotes x = -| and x = |, the graph of
y = arctanx has the corresponding horizontal asymptotes y - | and y = -~. And since arctanx is defined for all
x, the graph of y = arctan x has no vertical asymptotes.
2.61 Since sec x = 1/ cos x, and cos | = 0, sec | is undefined. Hence, | cannot be in the range of arcsec x.
28
Challenge Problems
2.62 Since tan 0 - , the equation 8 tan 0 = 3 cos 6 becomes = 3 cos 6, so 8 sin 9 = 3 cos2 9 = 3(1 - sin2 0).
Expanding and simplifying, we get 3 sin" 0 + 8 sin 0—3 = 0, which factors as (sin0 + 3)(3sin0-l) = 0, so sin 0 = -3
or sin 0 = y But -1 < sin 0 < 1 for all 0, so sin 0 cannot be equal to —3. Therefore, sin 9=1.
Challenge Problems
2.63
(a) First, we graph the functions y = sin(x - f) (solid) and y = cosx (dashed).
(b) The two graphs appear to be reflections of each other over the x-axis, which suggests that we have the
equation sin (x - - -cosx for all x.
2.64 The period of the function sin(yx + c\) is the same as the period of the function sinpx, which is But the
period of the graphed function is y, so ^ = y, which means \p\ = 4/3, so p = 4/3 or p = -4/3.
First, we take the case p = 4/3. The graph of y = sin y is shown below, which we must then translate to
produce the graph given in the problem.
The graph of y = sin y passes through the origin, and is increasing at the origin. The graph given in the
problem passes through the point ( |,0), and is increasing at this point. Hence, one possible equation for the given
graph is
4 ( n\
y = sin
.3 V 4).
However, since the sine function has period 2n, the equation
4 71 /4 (6n-l)7i\
y = sin -x - — + 2nn = sin
3 3 V3J+ "3 /
will also produce the same graph, for any integer n.
Next, we take the case p = -4/3. The graph of y = sin(-y) is shown below.
29
The graph of y = sin(-y) passes through the origin, and is decreasing at the origin. The graph given in the
problem passes through the point (-|,0), and is decreasing at this point. Hence, one possible equation for the
given graph is
r—4 ( n\
x+ —
.
= sm |
f 4 2n\
—
L 3 V 2/. ^ 3 3 J
Just as in the previous case, the equation
4 2tl 4 2(3n - l)n

3/ = sin — + Inn = sin \--x +
-3*
will also produce the same graph, for any integer n.
Therefore, the possible pairs (p, q) are
/4 (6n-l)n\ ^_4 2(3n-l)7tj

and
\3' 3 )
where n is any integer.
2.65 We claim that the period of /(x) = sin x cos x y

is 71.
First, we prove that cos(0 + 71) = - cos 0 and P = (cos

sin(0 + 71) = - sin 0 for all 0. Let P and Q be the
terminal points of 0 and 0 + n, respectively, so
P = (cos 9, sin 9) and Q = (cos(0 + 7i), sin(0 + n)).
The point Q is obtained by rotating P around

the origin counterclockwise by n, which means
that P and Q are diametrically opposite points
on the unit circle. Therefore, we have cost(9 + n) = cos 9 and sin(0 + 71) = - sin 0. Hence,
f(x + n) = sin(x + 71) cos(x + n) = (- sin x)(- cos x) = sin x cos x = /(x).
To show that n is the period of /(x), we must also show that there is no positive real number d smaller than n
such that /(x + d) = f(x) for all x.
If d is the period of /, then we have f(x + d) = 0 whenever f(x) = 0. So, finding all values of x such that f{x) = 0
will limit the possible values of d. We have /(x) = 0 if and only if cos x = 0 or sin x = 0, so we have /(x) = 0 for
x = y1 for all integers n. Since the values of x for which f(x) - 0 are | apart, the only positive real number d less
than n that could possible be the period of f(x) is f. However,
J_ J_ _ 1
V2 V2 _ 2'
and
. 371 371 1_ 1
= sin — cos —
4 4 Ti 2'
30
Challenge Problems
so /(f) ^ /(? + f)/ which means that the period of /(x) cannot be We conclude that the period of /(x) is [rT.
2.66 Since | sinx| < 1 for all x, if x satisfies the equation sinx = then
< 1 \x2\ < 6257T2 \x\ < V625ti2 = 25ti.

625ti2
Thus, all solutions x lie in the interval -2571 < x < 2571.
-2
We consider the graphs of y = sin x and y = and count the number of times they intersect. The graph of
y = is an upward-opening parabola. When |x| < 2571, we have < 1, so for |x| < 2571, the parabola intersects
the sine curve at two points in each period—once as sin x goes up from 0 to 1, and then again as sin x goes from 1
back down to 0. Therefore, in the 24 full periods from x = -247T to x = 24n, there are 2 • 24 = 48 intersections. We
have to be careful in the intervals —2571 < x < -2471 and 2471 < x < 2571. The two intersections in each period are
always in the left half of the period, since sin x is nonnegative in this half of the period. Therefore, there are two
intersections for 2471 < x < 2571, but no intersections for -2571 < x < -2471. This gives us a total of 48 + 2 50
intersections, and therefore 50 solutions to the given equation.
2.67 Let P be the terminal point of and let S be the foot of the perpendicular y
from P to the x-axis. Then sin ^ = PS, so arccos(sin = arccos PS.
But ZPOS = 7i - y = f, and ZOPS = f - f = if, so cos if = dl = Since ||

is in the range of the arccos function, we conclude that arccos PS = |f .
2.68 The terminal point of f is (cos f,sin f ) = (^/^), and the terminal V
point of ^ is (cos sin = (~:f /j)- Hence/ sin* > \ for f < x <
and sin x < \ for î<x<^+27t = ^|zi.
Since sinx has period 271, we can also say that sinx < \ for ^ + Ink <
x < i|n + 2nk for any integer k, which simplifies as 12n6fe-t5 < x < 12n^+13.
Then taking x = 30, we find that sinx = sin30, and 12n6-+- < 30 < Unk6+13.
Therefore, sin 30 < \ if and only if
12nk + 5 ~ link + 13
-^— < 0 <-—r-
for some integer k.
2.69
(a) We know that sin x and cos x have period 2n. Therefore,
g(x + 2n) = sin[fl(x + 2ti)] + cos[b(x + 2tt)] = sin(ax + 2nd) + cos(bx + 2nb) = sin ax + cos bx = g(x),
so g(x) is periodic. (However, the period of g(x) is not necessarily 2n. For example, the function sin2x has
period n.)
31
(b) Let a = «i/«2 and b = bi/bj, where a\, «2, h, and b2 are integers. Then aai - a\ and bbj = b\, and
g(x + 2TTA2&2) = sin[fl(x + 271^2^2)] + COS [b(x + 271^2^2)]

= sin(ax + 27iflfl2^2) + cos (bx + 271^2^2)
= sin(flx + 2na\b2) + cos (bx + Inaibi)

= sin ax + cos bx
= g(x),
so g(x) is periodic.
2.70 We consider the sine function with respect to radians. Then the sine of t in radians is simply sin t. However,
to evaluate the sine of t in degrees, we must first convert t to radians, which becomes ^t. Then the sine of t in
degrees is sin j^t. Since the professor found the correct answer despite being in the wrong angular mode, we
must have
sin( = sin—(.
In order for the sines of two angles to be equal, the terminal points of the two angles must have the same
y-coordinate. So, we examine the terminal points of the angles t and j^t.
For all positive t, the angle t is larger than the angle ^f. Therefore, when 0 < t < n, the terminal points of t
and t cannot be the same. Moreover, as t goes from | to n, the y-coordinate of its terminal point goes from 1
2 2
to 0. Meanwhile, the value of goes from to W- The y-coordinate of the terminal point of j^t therefore is
pretty close to 0, and increases slightly as t goes from f to n. Putting these together, we see that for some value of
t between | and tz, the two terminal points have the same y-coordinate.
Let these terminal points be P and Q, and let A and B be the feet of the altitudes
from these points, as shown at right. (This figure is very much not to scale!) Since
these terminal points have equal y-coordinates, we have QB = PA. Putting this
together with PO = QO gives us APOA = AQOB by SA Congruence for right trian¬
gles, so we have /QOB = /.POA. Since Q is the terminal point of /QOA, we have
/QOA - t. Similarly, we have /POA - ~~t, so we have
n = /QOB + /QOA - /POA + /QOA = -^-f + t.
Solving for t gives t -
Fortunately, finding the next solution is easier. As t continues from the value we just found up to t - n, the
y-coordinate its terminal point decreases, while that of ^ t increases, so there are no more positive solutions less
than or equal to n. For n <t < 2n, the value of sin t is negative, but that of sin is positive, since 0 < t < \ for
values of t in this range. The next time the angles t and j^t will have equal sines is when their terminal points
coincide. This occurs when the values of the angles themselves differ by 2n, so we have
n
t - 2n
Solving this equation gives t =
360ti 180ti
Therefore, the two smallest positive values of x such that sinx = sin ^x are 180-tt
and 180+ti
2.71 Since the price of the stock varies sinusoidally with time, we can express it as p(t) = a sin(bt + c) + d for some
constants a, b, c and d, where t is in days. Since the price per share varies from 80 to 120 dollars, the amplitude is
20, and the price varies about an average price of $100 per share. So, we can let a - 20 and d = 100, and we have
p(t) = 20 sin (bt + c) + 100.
32
Challenge Problems
At the start of the year, the price is 110 dollars, so we must have p(0) = 110, which means 20sin(c) = 10, so
sin(c) = i. Therefore, we can let c = and we have p(t) = 20 sin (bt + + 110. Finally, from the information that
the price climbs to a peak of 120 in 3 days, we have p(3) = 120, so 20 sin (3b + f ) + 100 = 120. From this, we have
sin (3b + |) = 1, so 3b + | which gives us b = |. Therefore, we have
p(0 = 20sinQt+ +100.
To maximize the amount of money we make, we buy as low as we can and sell as high as we can. So, we start
by buying at $110/share, since we know we'll be able to sell it at $120/share. We can buy 1000/110 shares, and we'll
sell them for $120 per share, so we'll have 120(1000/110) dollars.
After that, we repeatedly buy at $80 per share and sell at $120 per share, multiplying our money by 120/80 =
1.5 each time. But how many times can we do so? The period of p is ^ = 18. That is, stock goes from
$120/share to $80/share, and back up to $120 per share in 18 days. So, every 18 days, we can multiply our
money by 1.5, which means we'll be able to do so 360/18 = 20 times in the next 360 days, giving us a total of
120(1000/110) ■ 1.520 « $3,628,000. At that point, we'll be 363 days into the year, and the price will go down for
the last two days, and we won't be able to make any more money. Since we started with $1000, our profit then is
$3,628,000 - $1000 = $3,627,000
33
CHAPTER 3. TRIGONOMETRIC IDENTITIES
CHAPTER
I Trigonometric Identities
cos2 0 „ cos2 0 + sin2 0 1

3.1.1 cot2 0 + 1 —9— + 1 =-5- = esc2 0.
sin2 0 sin2 0 sin2 0
3.1.2 The angle 360° - 0 has the same terminal point as the angle -0. Let P and Q
be the terminal points of 0 and -0, respectively. The actions of rotating (1,0) an angle
0 counterclockwise about the origin and rotating (1,0) an angle 0 clockwise about the
origin are symmetric about the x-axis. That is, P and Q are reflections of each other
over the x-axis, which means that their y-coordinates are opposites. Therefore, we have
sin 0 = - sin(-0) = - sin(360° - 0), which means that sin(360° - 0) = - sin 0.
3.1.3 From the equation tan x + sec x = V3, we have sec x = V3 - tan x. Substituting into
the identity tan2 x + 1 = sec2 x, we get
2 _ V3
tan2 x +1 = (V3 - tan x)2 tan2 x +1 = 3 - 2 V3 tan x + tan2 x 2 V3 tan x = 2 tanx =
2V3 ~ 3 ‘
The only values in the interval 0 < x < 2n such that tan x = -^ are x = | and x = ^. Checking these values, we
find that only x= j satisfies the given equation.
3.1.4 We can combine the identities that we have derived so far. For example.
,'371 , n
cos I — + 0 I - cos L- + (71 + 0)
In the text, we showed that cos (f + x) - sin x for all x, so
~n
cos + (71 + 0) - sin(7i + 0),
.2
which we can rewrite as - sin[7i - (-0)].
We also showed that sin(7t - x) = sinx for all x, so -sin[7i - (-0) - -sin(-0). Finally, we know that
sin(-0) = - sin 0, so we have - sin(-0) = sin 0. Therefore, cos ( ^ + 0) = sin 0 .
3.1.5 First, we note that tan ^ = tan j = 1, and that we can pair off the other four angles conveniently:
n + if = 12 + 12 = 2 • -*n ^text'we showed that tan ( f - x) = cot x, so tan - x) = which means we have
34
Section 3.2
tan ^ tan ff = 1 and tan ^ tan ff = 1. Therefore, we have
n 2n 3n An 5n
tan — tan — tan — tan — tan —
12 12 12 12 12
3.1.6 The numerator is equal to
sin A(3 cos2 A + cos4 A + 3 sin2 A + sin2 A cos2 A) - sin A[(3 cos2 A + 3 sin2 A) + cos2 A cos2 A + sin2 A cos2 A]
= sinA[3 + cos2 A(cos2 A + sin2 A)]

= sinA(3 + cos2 A),
and, when A is not an integer multiple of the denominator is equal to
sinA / 1 . „ sinA sinA(l - sin2 A) sin A cos2 A

tanA(secA - sin A tan A) sinA.
cos2 A cos2 A
Therefore,
sinA(3 + cos2 A)
f(A) = sinA
3 + cos2 A.
The range of cos2 A is the interval [0,1], so the range of 3 + cos2 A is the interval [3,4], However, A is not allowed
to be any integer multiple of so cos A cannot be equal to 0,1, or -1. Therefore, the range of /(A) is the interval
(3,4)
3.2.1
(a) sin(7i + 9) = sin n cos 9 + sin 9 cos n = - sin 0
(b) cos + = cos | cos 9 - sin f sin 9 = -sin0
(c) cos(37i - 9) = cos 3n cos 9 + sin 3n sin 9 = cos 6

tan @+tan 4?
(d) If we apply the angle sum identity for tangent, we obtain tan (0 + ^) However, tan ^ is not
1-tan 0 tan y
defined, so we try a different approach:
tan 0+
3n sm 0 + t)
cos (e + f)'
By the angle sum identity for sine, we have
3n 3n . 3n
sin 0+ - sin 0 cos — + sm — cos 0 = - cos 0,
and by the angle sum identity for cosine, we have
3n 3n . 3n
cos 9+ = cos 0 cos —— sm 0 sin — = sm 0,
sin (0 + y) - cos 0
so -7-v =-= - cot 0 =
cos (0 + 3f) smd tan0
35
tan a + tan y
3.2.2 By the angle sum identity for tangent, we have tan(a + y) Letting y = -/3, we have
1 - tan a tan y
, tana + tan(-B) tan a -tan B
tan(a - B) = --
1 - tan a tan(-jS) 1 + tan a tan j3
3.2.3 By the angle sum identity for tangent, we have
tan 240° + tan 45° y/3 + 1 V3 + 1

tan 285° = tan(240° + 45°) =
1 - tan 240° tan 45° 1-V3-1 1-V5
(V5+ 1)(1 + V3) _ 4 + 2V3
(1 - V3)(l + V3) " "2
3.2.4 We have sin(a - jS) = sin a cos j3 - sin jS cos a and sin(a + j3) = sin a cos (i + sin cos a, so
sin(a - j8) sin(a + j3) = (sin a cos - sin fi cos a)(sin a cos + sin cos a)
= sin2 a cos2 j8 + sin « cos a sin j8 cos j8 - sin a cos a sin j8 cos jS - sin2 jS cos2 a
-- sin2 a cos2 - sin2 j3 cos2 a
= sin2 a(l - sin2 ft) - sin2 j3(l - sin2 a)
= sin2 a - sin2 a sin2 - sin2 j3 + sin2 a sin2
= sin2 a - sin2 j8.
3.2.5 If sin 2x sin 3x = cos 2x cos 3x, then
cos 2x cos 3x - sin 2x sin 3x - 0.
By the angle sum identity for cosine, we have cos 2x cos 3x - sin 2x sin 3x = cos(2x + 3x) = cos 5x, so cos 5x = 0.
Since x is an acute angle, we have 0<x< so 0 < 5x < The only angles between 0 and ^ whose cosine is 0
371
are \ and so the solutions are x = ^ and 10
3.2.6 By the angle sum identity for tangent, we have tan(A + B) = But we are given tan A + tanB =
1 - tan A tan B, so = 1/ which means tan(A + B) = 1.
Since angles A and B are acute, we have 0° < A + B < 180°. The only angle in this range whose tangent is 1 is
45°, so A + B = 45°. Thus, we have the equations A + B = 45° and A - B = 41°. Adding these two equations, we
get 2A = 86°, so A = f43°~
3.2.7 Squaring the given equations, we get
sin2 a + 2 sin a sin b + sin2 b =

3'
cos2 a + 2 cos a cos b + cos2 b = 1.
Adding these equations, we get
But sin2 a+cos2 fl = sin2b+cos2b = 1, so 2 sin a sin b+2 cos a cos b = |-2= |, which means cos a cos b+sin a sin b = |.
Then by the angle difference identity for cosine, we have cos(a - b) = cos a cos b + sin a sin b = i
36
Section 3.3
3.2.8
(a) Since 0 < a < | and y = | - a, we have 0 < y < Hence, y is acute. Similarly, 0 < /3 < \ and )]•
(c) Since sin(7i - x) = sin x for all x, we have sin(« + ft) - sin[7i - (y + cp)] = sin(y + cp).
(d) We see that
Y + (f)= (^~a) + (^~ P) =n-(a + p).
Since a + ft is obtuse, y + <p is acute. Furthermore, both y and <p are acute, which is the case we have already
proved (this is why it is important that y and <p are acute). So, we may apply the angle sum identity to get
sin(y + cp) - sin y cos <p + sin <p cos y.
But sin (| - x) = cos x and cos ( j - x) = sin x for all x, so
siny coscp + sincpcosy = sin ^ - a'j cos - P) + sin - jSj cos - ctj
= cos a sin j3 + cos j8 sin a.
By part (c), sin(a + ft) - sin(y + cp). Therefore, the angle sum identity for sine holds when a and /3 are acute
and a + jS is obtuse.
(e) If a and ft are obtuse, then n - a and 7Z - j8 are acute, so we can use the sine angle addition identity, which
we have proved for acute angles, to find that
sin ((71 - a) + (n - ft)) = sin(7t - a) cos(n - /3) + sin(7i - jS) cos(n - a).
Applying sin(7t - 0) = sin 0 and cos(7i - 0) = - cos 0, we have
sin ((71 - a) + (n - /?)) = - sin a cos - sin f3 cos a.
As Exercise 3.1.2, we showed that sin(27i - 6) - - sin 0, so
sin ((7i - a) + (tz - /?)) = sin (271 ~(a + ft)) = - sin(a + /J),
which means we have - sin(a+jS) = - sin a cos /3-sin ft cos a. Multiplying by-1 gives the desired sin(n+jS) =
sin a cos ft + sin ft cos a.
3.3.1 By the half-angle formula for sine, we have
45° /1 - cos45° /1 - V2/2 \l-^l

sin 22.5 = sin — = ± y-j-= = ± 2-'
Since the angle 22.5° lies in the first quadrant, its sine is positive, so sin 22.5° =
3.3.2 By the double angle formula for cosine, we have cos 20 = 1-2 sin2 0 = = 1 -
37
3.3.3 We have
2tan* = 2-gf 2 sin x cos x „ ,
—--j— = 2 sin x cos x,
1 + tan2 x 1 + sir|2+. cos2 x + sin x
COS2 X
and by the double angle formula for sine, we have 2 sin x cos x = sin 2x. Therefore, we have sin 2x = 1+tanTt' aS
desired.
3.3.4 We have
sin x cos x sin2 x + cos2 x 1 2
tan x + cot x =-b —— = —:-= —-= —-.
cosx sinx smxcosx smxcosx 2smxcosx
By the double angle formula for sine, we have 2 sin x cos x = sin 2x, so --= ——— = 2 esc 2x.
2 smxcosx sm2x
3.3.5 By the double angle formula for cosine, we have cos 2A = 2 cos2 A - 1 = 1-2 sin2 A, so
sin 2A - cos 2A + 1 _ sin 2A - (1 - 2 sin2 A) + 1 _ sin 2A + 2 sin2 A

sin 2A + cos 2A + 1 sin2A + (2 cos2 A - 1) + 1 sin 2A + 2 cos2 A
By the double angle formula for sine, we have sin 2A = 2 sin A cos A, so
sin2A + 2sin2A _ 2 sin A cos A + 2 sin2 A _ 2sinA(cosA + sin A) _ sin A

sin2A + 2cos2A 2sinAcosA + 2cos2 A 2cos A(sinA + cos A) cosA
3.3.6 Let x - cos 0 - sin 6. Then
x2 = cos2 6-2cos9sin9 + sin2 0 = 1-2cos0sin0.
By the double angle formula for sine, we have 2cos 0 sin 0 = sin20 = so x2 = 1 - ~ = ^, which means
x = ± = ±1. We are given that cos 0 > sin 0, so x = cos 0 - sin 0 = 2/5
3.3.7 Squaring the given equation, we get
3
cos2 0 + 2 cos 0 sin 0 + sin2 0 = -.
2
But cos2 0 + sin2 0 = 1, so 2cos 0sin 0 = If cos 0 is negative, then sin 0 is also negative, because the product
cos 0 sin 0 is positive. This means cos 0 + sin 0 is negative. But cos 0 + sin 0 = V3/2, so we cannot have cos 0 be
negative. Therefore, cos 0 is positive, so sin 0 is also positive, which means 0 lies in the first quadrant. In other
words, 0 < 0 < |.
Now, by the double angle formula for sine, we have 2 cos 0 sin 0 = sin 20, so 2 cos 0 sin 0 = \ gives us sin 20 = \.
We know that 0 < 0 < |, so 0 < 20 < n. The only angles in this interval whose sine is \ are ~ and Hence,
20 = | or 20 = so the solutions are 0 = yj and ||
We check these solutions: Note that | so by the angle difference identity for cosine.
7i 7T \
/ 7i n n . n . n 1 V2 V3 V2 _ V2+ V6
cos — = cos — —- = cos — cos — + sin — sm —
12 V3 4/ 3 4 34 2 ' 2 + 2 ' 2 4
and by the angle difference identity for sine.
71 . / 71 71 \ .77 71 . 77 77 V3 V2 V2 1 _ V6-V2
sm — = sm — - — = sm — cos - - sin - cos —
12 \3 4/ 34 43 2 ’ 2 2 ‘ 2 ~ 4
38
Section 3.3
n . n 2V6 V6 [6
so cos — + sm — =-= — = \ -
12 12 4 2 V 4
in
K
571 (n 57T\ 5n . n n
• (71
cos — + sm — = cos -x- -— + sin — sm — + cos —
12 12 '^2 12/ V 2 " 12 12 12
3.3.8
(a) By the angle sum identity for cosine, we have
cos 30 = cos(0 + 20) = cos 0 cos 20 - sin 0 sin 20.
Then by the double angle formulas for cosine and sine, we have
cos 0 cos 20 - sin 0 sin 20 = cos 0(2 cos2 0 - 1) - sin 0(2 sin 0 cos 0) = 2 cos3 0 - cos 0-2 sin2 0 cos 0.
But sin2 0 = 1- cos2 0, so
2 cos3 0 - cos 0-2 sin2 0 cos 0 = 2 cos3 0 - cos 0 - 2(1 - cos2 0) cos 0
= 2 cos3 0 - cos 0-2 cos 0 + 2 cos3 0
= 4 cos3 0-3 cos 0 .
(b) By the angle angle sum identity for sine, we have
sin 30 = sin(0 + 20) = sin 0 cos 20 + sin 20 cos 0.
Then by the double angle formulas for cosine and sine, we have
sin 0 cos 20 + sin 20 cos 0 = sin 0(1 - 2 sin2 0) + (2 sin 0 cos 0) cos 0
= sin 0-2 sin3 0 + 2 sin 0 cos2 0.
But cos2 0 = 1- sin2 0, so
sin 0 - 2 sin3 0 + 2 sin 0 cos2 0 = sin 0 - 2 sin3 0 + 2 sin 0(1 - sin2 0)

= sin 0 - 2 sin3 0 + 2 sin 0-2 sin3 0
= 3 sin 0-4 sin3 0 .
(c) By part (a), cos 3x = 4 cos3 x - 3 cos x, so
cos3x 4 cos3 x - 3 cos x . 2

-=-= 4 cos x - 3.
cos x cos x
Hence,4cos2x-3 = l/3,socos2x = (3 + l/3)/4 = 10/12 = 5/6. Therefore,sin2x = l-cos2x = 1-5/6 = 1/6.

By part (b), we have sin 3x = 3 sin x - 4 sin3 x, so
sin 3x 3 sin x - 4 sin3 x 9 1

= 3-4 sin2 x = 3 - 4 • -
sinx sinx
39
cos 4x - cos 2x -2 sin 3x sin x

3.4.1 By the difference-to-product formula for cosine, we have sinx.
2 sin 3x 2 sin 3x
3.4.2 By the sum-to-product formula for cosine, we have cos 27° + cos 33° = 2 cos 30° cos 3° V3cos3°, so
0 =
3.4.3 Subtracting the second equation from the first equation and dividing by 2, we get
sin jS cos a = -[sin(n + jS) - sin(n - jS)].
Switching a and /3 and noting that sin(jS - a) = sin(-(n -/?)) = - sin(n - j3), we get
1 1
sin a cos /? = - [sin(n + jS) - sin(jS -«)] = - [sin(n + jS) + sin(n - j8)],
which agrees with our earlier formula.
3.4.4 By the sum-to-product formula for sine, we have
sin 10° + sin 80° = 2 sin 45° cos 35° = V2 cos 35°,
sin 20° + sin 70° = 2 sin 45° cos 25° = V2cos25°,
sin 10° + sin 20° + sin 30° + sin 40° + sin 50° + sin 60° + sin 70° + sin 80°
= V2(cos5° + cos 15° + cos 25° + cos 35°).
By the sum-to-product formula for cosine, we have
cos 5° + cos 35° = 2 cos 20° cos 15°,

cos 15° + cos 25° = 2 cos 20° cos 5°,
so
V2(cos5° + cos 15° + cos 25° + cos35°) = V2(2 cos 20° cos 15° + 2 cos 20° cos 5°) = 2 V2 cos 20°(cos 5° + cos 15°).
Finally, by the sum-to-product formula for cosine, we have cos 5° + cos 15° = 2 cos 10° cos 5°, so
2 V2 cos 20° (cos 5° + cos 15°) = 4 V2 cos 5° cos 10° cos 20°.
Therefore,
sin 10° + sin 20° + sin 30° + sin 40° + sin 50° + sin 60° + sin 70° + sin 80° 4 V2 cos 5° cos 10° cos 20° \~T~nT
— T VZ
cos 5° cos 10° cos 20° cos 5° cos 10° cos 20°
3.4.5 By the difference-to-product formula for sine,
sin(B + C - A) - sin(A + B + C) = 2 sin(-A) cos(B + C) -2 sinAcos(B + C),
40
Section 3.4
and by the sum-to-product formula for sine,
sin(C + A - B) + sin(A + B - C) = 2 sin A cos(C - B).
Hence,
- sin(A + B + C) + sin(B + C - A) + sin(C + A — B) + sin(A + B — C) = — 2 sin A cos(B + C) + 2 sin A cos(C - B)

= 2sinA[cos(C - B) - cos(B + C)].
By the difference-to-product formula for cosine.
cos(C - B) - cos(B + C) = -2 sin C sin(-B) = 2 sin B sin C,
so 2 sin A [cos(C - B) - cos(B + C)] = 4 sin A sin B sin C.
3.4.6 The sum in the numerator and difference in the denominator suggest we should try the sum-to-product
and difference-to-product identities, particularly because the angles are the same, so the difference of the angles
is 0°, which we know how to handle. Unfortunately, all these identities involve the sum or difference of two of
the same trig function; none of them has a sum or difference of one sine and one cosine. So, we try changing the
problem into a form that involves the sum or difference of two sines or the sum or difference of two cosines.
There are many ways to turn a sine into a cosine. Clearly, sin2 6 + cos2 0 = 1 isn't what we're looking for, since
this won't give us a sum of cosines. But sin 0 = cos(90° - 0) will. Applying this to both of the sines in the problem
gives
cos 96° + sin 96° _ cos 96° + cos(-6°)
cos 96° - sin 96° cos 96° - cos(-6°) ’
We then apply the sum-to-product and difference-to-product identities for cosine:
cos 96° + cos(-6°) 2cos(S^1)cos(^2) 2 cos 45° cos 51°

= - cot 51°,
cos 96° - cos(-6°) 2 sin 45° sin 51°
so the equation we must solve now is tan 19x° = - cot 51°. We'd rather have tangent on both sides, so we convert
cotangent to tangent. Since
cos 51° sin(90° - 51°)

cot 51° = tan(90° - 51°) = tan 39°,
sin 51° cos(90° - 51°)
we can write tanl9x° = -cot51° as tanl9x = - tan39°. We'd like to get rid of the negative sign so that we're
equating two tangents. We have plenty of identities to do that. The two simplest are tan 0 = - tan(-0) and
tan0 = -tan(180° - 0). The former gives tanl9x° = -tan39° = tan(-39°) and the latter gives tanl9x° =
-tan39° = tan 141°. Uh-oh! These look different! But they're not. Since the period of tan0 is 180°, we have
tan(-39°) = tan(-39° + 180°) = tan 141°. This observation also helps us see how to finish the problem.
We want tan 19x° = tan 141°, but we also need x to be a positive integer. If we let 19x = 141, then x is not
an integer. But we also know that tan 141° = tan(141° + 180n°) for all integers n. So, we keep increasing 141° by
multiples of 180° until we hit a multiple of 19°, which will allow us to divide by 19° to get an integer answer for x.
Now we have a basic number theory problem. We want the smallest positive integer x such that there is some
positive integer k with 19x = 141 + 180k. Dividing by 19 gives
8 9k 9k+ 8
x = 7+ — +9k+ — =9k + 7 +
19 19 19
We could figure this out with trial and error, starting with k-1 and increasing k by 1 until 9k + 8 is divisible by
19. Instead, we'll use a little more clever algebra and number theory.
41
Since 9k + 7 is an integer, we must have (9k + 8)/19 be an integer in order for x to be an integer. If (9k + 8)/19 is
an integer, then there is a positive integer / such that 9k + 8 = 19/'. Solving for k gives
19 j - 8
k=
9
We want k to be an integer, so we let / = 8, since this is the smallest possible positive integer value of / that makes
(/ - 8)/9 an integer. When j = 8, we have k = 16, which gives
9k+ 8
x = 9k + 7 + 159
19
3.5.1 From the identity cos0 = sin(90° - 0), we have cos41° = sin49°, so cos41° + sin41° = sin49° + sin41°.
Then by the sum-to-product formula for sine, we have
sin 49° + sin 41° = 2 sin 45° cos 4° = 2 • • cos 4° V2cos 4°.
Hence, we have cos41° + sin41° = V2cos4°. Comparing this to the given cos41° + sin41° = V2sinA gives
sin A = cos 4°. Since cos 4° = sin(90° - 4°) = sin 86°, we have A = 86°
3.5.2 We let ad + be = k. We square this equation and the equation ac - bd = hoping to produce squares that
we can combine with our other two equations. These squarings give us
a2d2 + labed + b2c2 = k2,

l
a2c2 - 2abed + b2d2 = -.
4
Adding these two equations eliminates the abed terms and leaves a2d2 + b2c2 + a2c2 + b2d2 = k2 + |. The left
side of this equation is exactly what we get when we multiply the left sides of the other two equations. Since
(a2 + b2)(c2 + d2) - 1 we have
1 = a2c2 + a2d2 + b2c2 + b2d2 = k2 +
which means k2 = |. Since a, b, c, and d are positive, k must also be positive, and we have k = 2
3.5.3 We can solve this problem with sin2 0 + cos2 0 = 1, but it's a little tricky—try it and see. Instead, we'll use
the sum-to-product identities. Applying these identities gives
A+B A-B
2 sin cos
2
„ :a + b A-B
2 cos —-— I cos
We can eliminate the terms with (A - B)/2 by dividing the first equation by the second (which we can safely do
since we know the left side of the second equation is not 0). This gives us
42
Section 3.5
The angles are the same on the left side, so we can write this equation as tan ( ) = V3. Since we seek the
value of A + B such that 0° < A, B < 180°, we know that 0 < < 180°, and the only angle in this range whose
tangent equals V3 is 60°. Therefore, we have = 60°, so A + B = 120 °
3.5.4 By the angle sum identity for tangent, we have tan(x + y) = yitln^tany * We are giyen that tanx + tan y = 25,
so we only need to find tan x tan y.
tan x+tan y
We are also given that cotx + cot y = 30. But cotx + cot y = -T- + -— sn
° J " tanx tany tan x tan y '
tan x + tan y _ 25 5
tan x tan y
cot x + cot y 30 6 '
Therefore, tan(x + y) = ta^-r+t^ny _ 150

’ v “/ 1-tan x tany 1-5/6
3.5.5 First, we determine cos B by noting that cos2 B = 1 - sin2 B = l- ^ = ^,so cos B = ± yj ^ = ±-^- We are
given that B is acute, so cos B - ^=, which means tanB = = |.
By the double angle formula for tangent, we have tan2B = - \• Since B is acute, we have
0° < 2B < 180°. Since tan2B is positive, 2B is also acute.
Then by the angle sum identity for tangent, we have
tan A + tan2B 1/7+ 3/4

tan(A + 2 B) =
1 - tan A tan 2B 1-1/7-3/4
Since both A and 2B are acute, we have 0° < A + 2B < 180°. The only angle in this range whose tangent is 1 is 45°,
so A + 2B = [45°
3.5.6 Rearranging the terms, we get sin 6 - cos 6 - cos 26 - sin 20. Squaring both sides, we get
sin2 0-2 sin 0 cos 0 + cos2 0 = cos2 20-2 cos 20 sin 20 + sin2 20.
But sin2 0 + cos2 0 = cos2 20 + sin2 20 = 1, so the equation above becomes 1-2 sin 0 cos 0 = 1-2 cos 20 sin 20,
which means sin 0 cos 0 = cos 20 sin 20.
By the double angle formula for sine, we have sin 26-2 sin 0 cos 0, so
sin 0 cos 0 = 2 cos 20 sin 0 cos 0.
Moving all the terms to one side, we get
2 cos 20 sin 0 cos 0 - sin 0 cos 0 = 0,
which factors as
sin 0 cos 0(2 cos 20 - 1) = 0.
Hence, sin 0 = 0, cos 0 = 0, or cos 20 = We are given that 0 is acute, so neither sin 0 nor cos 0 can be equal to 0.
Therefore, cos 20 = \. Since 0 is acute, we have 0 < 20 < 7t. The only angle in this range whose cosine is \ is |, so
_ 1 , V3
20 = ?, and 0 = . This solution works, because when 0 = f,we have sin 0 + sin 20 = cos 0 + cos 20=1 +
3.5.7 By the angle sum identity for cosine and the angle difference identity for sine, we have
cos(a + fi) = cos a cos ft - sin a sin p,

sin(a - P) = sin a cos j3 - cos a sin p,
43
so the given equation becomes cos a cos p - sin a sin p + sin a cos p - cos a sin p = 0. Dividing by cos a cos p, we get
sin a sin B sin a sin B

1-T + —-+ = 0,
cos a cos p cos a cos p
which becomes 1 - tan atan p + tan a - tan p = 0. This equation factors as (1 + tan a)(l - tan p) = 0, so tan a = -1
or tan p = 1. We are given that tan p = 1/2000, so tan a = | -11.
3.5.8 Expanding the given equation, we get
1 + sin t + cos t + sin t cos t = -.

4
Let x = (1 - sin f)(l - cos f), so

1 - sin t - cos t + sin t cos t = x.
4,
Adding these equations, we get 2 + 2 sin t cos t = x + |, so sin t cos t = *+5/4~2
2 = fcA.
8 Taking the difference of the
same two equations gives 2 sin t + 2 cos t = | - x, so sin t+ +
5 _ v _L
cos t* =
_ 5/4—JC _ 5-4y
Squaring this equation, we get
5 - 4x
sin21 + 2 sin t cos t + cos2 t =
8
However, sinz t + cos2 t = 1 and sin t cos t = (4x - 3)/8, so
4x - 3 ( 5 - 4x
1+2
8 8
Simplifying, we obtain the quadratic equation 16x2 - 104x + 9 = 0. By the quadratic formula, we have
+ Vl042 -4-16-9
104 104 ± 32 VlO 13 r—
-2+6-- 32 -T±VI°-
However, -1 < sin f < 1 and -1 < cos f < 1, so
x = (1 — sin f)(l - cos f) < 2 • 2 = 4,
and p + VlO > ^ + 3 > 4. Hence, we have x = f - VlO

3.5.9 Solving for y, z, and x in the given equations, respectively, we find
2x 2y 2z
V =
x =
l-x2' 1-3/2' 1 -z:2 1
The right sides of these equations look like the double angle identity for tangent. Since x, y, and z are real
numbers, we can let a = arctanx, p = arctany, and y - arctanz, so x = tana, y = tanp, and z = tany. Then from
y= / we have
„ 2 tan a
tanp = --
1 - tan2 a
But from the double angle formula for tangent, we have
2 tan a
—-—=— = tan 2a,
1 - tan2 a
so tan p = tan 2a. Similarly, from the other two initial equations, we have tan y = tan 2p and tan a = tan 2y.
44
Review Problems
We'd like to combine these equations in a way that allows us to produce an equation with only a. One tempting
possibility is to combine tan (i - tan 2a and tan y = tan 2/3 through the following argument: "Since tan /3 = tan 2a,
we have tan 2/3 = tan 4a, and combining this with tany = tan2jS gives tany = tan 4a." But before we get too
excited, we better check the claim that if the tangents of two angles are equal, then the tangents of double the
angles are equal.
To do so, we let 0 and i/> be two angles. Then
2tan0 2 tan1
tan 29 = and tan2t/>
1 - tan2 0 1 - tan2 ip ’
Hence, if tan 0 = tan tp, then tan 26 = tan 2i\>.
Returning to our problem, we know that tany = tan2/1, so tan2y = tan4/3. Also, tan/3 = tan 2a, so tan2/3 =
tan 4a, and tan 4/3 = tan 8a. Therefore,
tana = tan2y = tan 4/3 = tan 8a.
In particular, tana = tan8a. Two angles with equal tangents must differ by an integer multiple of n. To see why,
suppose that tan 6 = m. Then, by the definition of the tangent function, the terminal point of 6 is one of the two
points at which the line y = mx intersects the unit circle. These two points are at the opposite ends of a diameter
of the unit circle. That is, they are terminal points of angles that differ by an integer multiple of n.
Therefore, from tan a = tan 8a, we know that 8a = a + nk, where k is an integer. Solving for a, we find a = y,
so x = tan a - tan y.
We now check that this solution works. If x = tan y, then
2x 2 tan y 2nk
= tan
1 - X2 1 - tan2 y ~7~’
2y 2 tan y^ Ank
= tan
_ 1 -y2 ~ 1 - tan2
and
2z 2 tan ^y 8 nk
1 - z2 1 - tan2 ^ ‘
But ^ ^ ^ = nk, so tan = tan y = x. Hence, x = and the system is satisfied.
Therefore, the solutions are of the form
nk 2nk Ank
(x,y,z) = (tany, tan , tan ——
where k = 0,1,..., 6. (We only take k up to 6 because the solutions for k = 7 and k = 0 are the same, as are the
solutions for k = 8 and k — 1, and so on.)
Review Problems
sin(7i - 6) sin0
3.33 tan(7i - 6) = - tan 6.
cos(7T - 0) - cos 6
3.34 Since sin(-x) = - sinx for all x, we have sin(0 n) = - sin(7z - 0). We also have sin(n - x) = sinx for all x,
so sin(0 — n) = — sin(7i - 0) = - sin 0
45
3.35
(a) First, we have cos(x - y) - sin[90° — (x - y)] = sin(90° - x + y). Therefore, we have
cos(x - y) = sin(90° - x + y) = sin[(90° - x) + y] = sin(90° - x) cos y + sin y cos(90° - x).
Letting z = 90° - x in the identity sin(90° - z) = cos z gives sin x = cos(90° - x), so
cos(x - y) = sin(90° - x) cos y + sin y cos(90° - x) = cos x cos y + sin x sin y.
(b) Taking y = x in sin(x + y) = sin x cos y + sin y cos x, we get sin 2x = 2 sin x cos x.
(c) First, cos(x + y) = sin[90° - (x + y)] = sin(90° - x - y). Then
sin(90° - x - y) = sin[(90° - x) + (-y)] = sin(90° - x) cos(-y) + sin(-y) cos(90° - x) = cos x cos y - sin x sin y.
Fience, we have
. , . . sins , siny
sin(x + y) _ sin x cos y + sin y cos X _ cosx cosy tan x + tan y
tan(x + y)
cos(x + y) cosxcos y - sinxsin y i _ sm*smy 1 - tan x tan y
J j j cosxcosy
(d) Letting y = -x in the identity we proved in part (a), we have
cos 2x = cos x cos(-x) + sin x sin(-x) = cos2 x - sin2 x = cos2 x - (1 - cos2 x) = 2 cos2 x - 1.
Letting x = |, we have cosz = 2 cos2 | - 1, so cos2 | = 1+c2osz. If 0 < z < n, then 0 < | < |, so cos | > 0.
Therefore, we take the positive square root of our equation above to find cos | = y 1+--2—z.
(e) If part (a), we found
cos(x - y) = cos x cos y + sin x sin y.
Replacing y with -y in this equation, we have
cos(x + y) = cos x cos(-y) + sin x sin(-y) = cos x cos y - sin x sin y.
Adding these two equations gives cos(x - y) + cos(x + y) = 2 cos x cos y, so we have the identity cos x cos y =
|(cos(x + y) + cos(x - y)).
(f) Replacing y with -y in sin(x + y) = sin x cos y + sin y cos x gives
sin(x - y) = sin x cos(-y) + sin(-y) cos x = sin x cos y - sin y cos x.
Subtracting this from our equation for sin(x + y) gives
sin(x + y) - sin(x - y) = 2 sin y cos x.
Letting a = x + y and /3 = x - y, we then have x = (a + fi)/2 and y = (a - j3)/2, and we have sin a - sin jS =
2 sin ( ) cos ( ).
3.36 By the angle sum formula for sine, we have sin(x - y) cos y + cos(x - y) sin y = sin[(x - y) + y] sinx
3.37 Applying the angle difference identity for cosine, we have
V2 V2 . N
V2cos (^ _ = ^2 (cos ^ cosx + sin ^ sinxj = V2 —- cos x + —— sin x = cos x + sin x.
2 2
which is choice
46
Review Problems
3.38 Since cos2 0 = 1 — sin2 0, we have
-3 + 4 cos2 0 -3 + 4(1 - sin2 0) 1-4 sin2 0 (1 + 2 sin 0)(1 - 2 sin 0)

1 + 2 sin 0.
1-2 sin 0 1-2 sin 0 1-2 sin 0 1—2 sin 0
Thus, a = 1 and b = 2
3.39 We know how to find the range of a function of the form f(t) = a sin t, so we try to write g(t) in this form.
The angle sum identity for sine has a single sine on one side and a sum of trig expressions on the other:
sin(f + y) = sin t cos y + sin y cos t.
However, we have g(t) = 2 sin t + 3 cos t, and there is no angle y such that cos y = 2 and sin y = 3. But maybe we
can find some constant c such that there is an angle y with cos y = 2 and sin y = ~c. From cos2 y + sin2 y = 1, we
have ^ + ^ = 1, from which we find that c = Vl3 fits the bill. That is, we have
2 .
—= smt + cos t) = Vl3 sin(f + y).
Vl3 Vl3
where y is the angle such that cosy = -^= and siny = -^=. (We know that such an angle exists because
(^fe)2 + (^fe)2 = 1' which means (^=, ^j=) is on the unit circle.)
From g(t) = Vl3 sin(t + y), where y is some constant angle, we see that the range of g(t) is - Vl3, Vl3 ,
because the range of sine is [-1,1].
3.40 We have
tan2 20° - sin2 20° sin2 20° _ sin2 20° - sin2 20° cos2 20° _ sin2 20° (1 - cos2 20°) _ sin4 20°
tan2 20° sin2 20° sin^20° . sin2 20° sin4 20° sin4 20° sin4 20°
cos2 20°
tan(a - (3) + tan j5

3.41 By the angle sum identity for tangent, we have tan a - tan[(a -/?) + /?]
1 - tan(« - jS) tan jS'
3.42 By the angle sum formula for tangent, we have
tan y + tan 0
tan
1 - tan y tan 0
Uh-oh; tan y is not defined! To get around this, we write tangent in terms of sine and cosine:
sin (f + 0) sin y cos ^ + cos ^ sin 0 - cos 0 1

tan - cot 0 =
cos (f + 0) cos y cos ^ - sh1 ^2 sin ® S4n ^ tan0
3.43 By the angle sum identity for cosine, we have cos(A + B) = cos A cos B - sin A sin B - 1. Because A and B
are angles of a triangle, we must have 0° < A + B < 180°. However, there is no angle in this range whose cosine
equals 1, so it is impossible for any A ABC to satisfy the equation cos A cos B - sin A sin B = 1.
3.44 By the angle sum formulas for sine and cosine, we have
V3 1
sin(;t + 30°) = sin x cos 30° + cos x sin 30° = — sinx + - cosx,
V3 1
cos(x + 30°) = cosxcos30° - sinxsin30° = — cos* - - sinx.
47
Then
V3 sin(x + 30°) - cos(x + 30°) V3 ( 4^ sin x + \ cos x) - ( 4^ cos x - \ sin x)

4 cos x sin(x + 30°) - 4 sin x cos(x + 30°) 4cosx (4p sinx + \ cosx) - 4sinx (4^ cosx - \ sinx)
2sinx
2 cos2 x + 2 sin2 x
sinx
3.45 We have arccos (—|) = 120° and arcsin = 45°, so
1 V5\
tan ârccos ( - + arcsin j = tan(120° + 45°).
tan 120° + tan 45° - V3 + 1 1 - V3 (1 - V3)(l - V3)

tan(120° + 45°) - V3-2
1 - tan 120° tan 45° 1 - (- >/3) • 1 1 + V3 (1 + y/3)(l - V3)
3.46 We know that
V6- V2
sin 15° = sin(45° - 30°) = sin 45° cos 30° - sin 30° cos 45° and
4
V6 +V2
sin 75° = sin (45° + 30°) = sin 45° cos 30° + sin 30° cos 45°
V6' 216 27
so (sin 15° + sin 75°)6 =
~64 ~8 ~
3.47 We have
sin 2x cos 2x sin 2x cos x - sin x cos 2x
sin x cos x sm x cos x
By the angle difference identity for sine, we have
sin 2x cos x - sin x cos 2x = sin(2x - x) = sin x,
sin 2x cos x - sin x cos 2x sinx 1

so secx
sm x cos x sm x cos x cos x
O
3.48 Since cos2 6 + sin 6 - 1, we have
,
sin2 0 = 1- cos2 0 = 1 - (-0.6)2 = 1 - 0.36 = 0.64,
so sin 0 = ± V0.64 = ±0.8. Then by the double angle formula for sine, we have sin 20 = 2 sin 0 cos 0 = ±0.96.
Now we show that both 0.96 and -0.96 are achievable. There exists an angle 0 in the second quadrant such
that cos 0 = -0.6 and sin 0 = 0.8 (since (-0.6)2 + (0.8)2 = 1). For this angle, we have sin 20 = 2 sin 0 cos 0 = -0.96.
There's an angle 0 in the third quadrant for which cos 0 = -0.6 and sin 0 = -0.8. For this angle, we have
sin 20 = 2 sin 0 cos 0 = 0.96. Thus, both 0.96 and -0.96 are possible values of sin 20.
3.49 We have
COS X _ 1
cot x -1 sin x cos x - sin x
cot x + 1 ± 1 cos x ± sin x'
48
Review Problems
Looking at the other side of the identity we hope to prove, we have
cos 2x _ cos2 x - sin2 x _ (cos x - sin x)(cos x + sin x)

1 + sin 2x 1+2 sin x cos x 1 + 2 sin x cos x
Aha! The factored numerator gives us a clue how to tie these together. We multiply the numerator and
denominator of our fraction for by cos x + sin X/ anj we j
cosx — sinx _ (cosx - sin x) (cos x + sinx) cos2 x-sin2 x cos2x cos2x
cosx + sinx (cosx + sinx)(cosx + sinx) cos2x + 2sinxcosx + sin2x ~~ 1 + 2sinxcosx l + sin2x'
where we have also used the identity cos2 x + sin2 x = 1 to simplify the denominator.
3.50 We start with the identity sin2 x + cos2 x = 1. Squaring both sides, we get
sin4 x + 2 sin2 x cos2 x + cos4 x = 1.
By the double angle formula for sine, we have sin2x = 2 sinx cosx. But we are given that sin2x = ||, so
2sinxcosx = Then sinxcosx = so 2sin2xcos2x = 2 (Therefore, sin4x + cos4x = 1 - fff = .
3.51 Applying the half-angle formula for tangent, we have
571 sin ^ £ -V2 _ -V2(2 + V2)

2
-1 - V2
tanT = 1 + cos ^ VI
2
2 - V2 “ (2 - V2)(2 + V2)
3.52 By the double angle formula for sine, we have sin 2x = 2 sin x cos x, so the inequality sin 2x < sin x becomes
sinx - 2 sinx cosx > 0, which means sinx(l - 2 cosx) > 0. This inequality is satisfied if and only if sinx > 0 and
cos x < or sin x < 0 and cos x > \.
For 0 < x < 271, sin x > 0 if and only if 0 < x < n, and cos x if and only iff <x<f. Hence, sin x > 0 and
cos x < \ if and only if f <x< 7i.
Similarly, sin x < 0 if and only if n < x < 2n, and cos x > \ if and only if0<x<for^<x< In. Hence,
sin x < 0 and cos x > \ if and only if ^ < x < In.
Therefore, the solution is x £ [f,77]U[f,27l]
3.53 Since cos2 9-1- sin2 9, the given equation becomes
2(sin2 9 - 1 + sin2 9) = 8sin0 - 5 <=> 4 sin2 9 - 8 sin 9 + 3 = 0 «=> (2 sin 9 - l)(2sin0 - 3) = 0,
so sin 9 = \ or sin 9 -
571
Since -1 < sin 9 < 1 for all 9, sin 9 can never be equal to so sin 9 - \. The solutions are 9 = f and 6
3.54
(a) By the double angle formulas for sine and cosine, we have sin 29 - 2 sin 9 cos 9 and cos 29 = 2 cos2 9 -1, so
sin 9 + sin 29 _ sin 9 + 2 sin 9 cos 9 _ sin 9 + 2 sin 9 cos 9

1 + cos 9 + cos 29 1 + cos 9 + 2 cos2 0-1 cos 0 + 2 cos2 0
sin 0(1 + 2 cos 0) sin 0 _
=-— =-- = tan 0.
cos 0(1 + 2 cos 0) cos 0
49
(b) Our derivation in part (a) fails if 1 + 2 cos 6 = 0, or cos 6 = -1/2. The only value in the interval \ < 6 < n
for which this occurs is 6 = y .
3.55 By the product-to-difference identity, we have
1
2 sin68° sin52° = cos(68° - 52°) - cos(68° + 52°) = cos 16° - cos 120° = cos 16° + -,
so we want to find the acute angle cp such that sirup = cos 16° = sin(90° - 16°) = sin 74°. Hence, <p =
3.56 To prove the identity, we express all the trigonometric functions in terms of sin x and cos x. The left-hand
side becomes
-tanx _|_ -cotx — -sin x/ cos x _|_ -cos x/sin

--
x
1-cotx 1-tanx 1-cosx/sinx 1-sinx/cosx
sin2 x cos2 x
cos x(sin x - cos x) sin x(cos x - sin x)
sin3 x cos3 x
sin x cos x(sin x - cos x) sin x cos x(sin x - cos x)
sin3 x - cos3 x
sin x cos x(sin x - cos x)'
As a difference of cubes, we have sin3 x - cos3 x = (sin x - cos x)(sin2 x + sin x cos x + cos2 x), so
sin3 x - cos3 x _ (sin x - cos x)(sin2 x + sin x cos x + cos2 x)

sin x cos x(sin x - cos x) sin x cos x(sin x - cos x)
_ sin2 x + sin x cos x + cos2 x
sin x cos x
1 + sin x cos x 1
= —;-= —-+1 - sec x esc x + 1,
sm x cos x sin x cos x
which is the right-hand side of the identity we wish to prove.
3.57 From the sum-to-product formulas, we have
sin x + sin 5x = 2 sin 3x cos 2x,

cos x + cos 5x = 2 cos 3x cos 2x,
sin x + sin 3x + sin 5x _ 2 sin 3x cos 2x + sin 3x _ sin3x(2cos2x + 1) sm 3x

-— = tan 3x.
cos x + cos 3x + cos 5x 2 cos 3x cos 2x + cos 3x cos3x(2cos2x +1) cos3x
Challenge Problems
3.58 Squaring the given equation, we get sin2 x + 2 sinxcosx + cos2 x = A. But sin2 x + cos2 x = 1, so we have
2 sinxcosx = ^ - 1 = -||, which means sinxcosx = Substituting cosx = g - sinx (this is a rearrangement
of the given equation) into sin x cos x = - gives
• /I . \ 12
Sm x \ 5 ~ sm x ) = - 25'
50
Challenge Problems
Multiplying both sides by 25, expanding the left side, and rearranging, gives 25 sin2 x - 5 sin x - 12 = 0. Factoring
the left side as a quadratic in x gives (5 sin x - 4)(5sinx + 3) = 0. Since we are given | < x < tl, we have sinx > 0,
and the only positive solution to the quadratic for sinx is sinx =
From cos2 x = 1 — sin2 x = ^ and the fact that x is in the second quadrant, we have cos x = -|, so
sinx
tan x =-
cosx
3.59 To find /(sec2 0), we find the value of x such that ~ = sec2 0. Taking the reciprocal of both sides, we get
TT ~ which becomes 1 - \ = cos2 0. Hence, we have
- = 1 - cos2 0 = sin2 0,
x
so /(sec2 0) =/(j=T) sin2 0
3.60 Since tan4x = the given equation becomes Cross-multiplying, we get the equation
cos x sin 4x + sin x sin 4x = cos x cos 4x — sin x cos 4x, so
cos x sin 4x + sin x cos 4x = cos x cos 4x - sin x sin 4x.
By the angle sum formula for sine, we have cos x sin 4x + sin x cos 4x = sin(x + 4x) = sin 5x, and by the angle
sum formula for cosine, we have cos x cos 4x - sin x sin 4x = cos(x + 4x) = cos 5x, so the equation above becomes
sin5x = cos5x. Then tan 5x = cos 5x = 1.
The smallest positive angle whose tangent is 1 is 45°, so the smallest possible positive angle x is 45°/5 =
3.61 To find the intersections of the graphs, we must solve the equation
8 cos x + 5 tan x = 5 sec x + 2 sin 2x.
By the double angle formula for sine, we have sin2x = 2 sin x cos x, so the equation above becomes
„ 5sinx 5 , .
8 cos x H-=-+ 4 sm x cos x
cos x cos x
=> 8 cos2 x + 5 sin x = 5 + 4 sin x cos2 x
o 8(1 - sin2 x) + 5 sin x = 5 + 4 sin x(l - sin2 x)
<=> 8-8 sin2 x + 5 sin x = 5 + 4 sin x - 4 sin3 x
<=> 4 sin3 x - 8 sin2 x + sin x + 3 = 0.
We can factor the left side of the final equation as a cubic polynomial. To see why, let t = sinx, so we have
4f3 - St2 + t + 3 = 0. Clearly, t - 1 is one solution, and factoring out t - 1 gives us (f - 1)(4f2 - 4f - 3) =
0, from which we have (t - l)(2f - 3)(21 + 1) = 0. Substituting t - sinx back into this equation, we have
(sin x — 1)(2 sin x - 3)(2 sin x + 1) = 0.
Hence, sinx - 1, sinx = |, or sinx = -\.
If sinx = 1, then cos2x = 1 - sin2x = 1 - 1 = 0, so cosx = 0. But if cosx = 0, then tanx = sinx/cosx and
sec x = 1 / cos x are undefined, so there are no solutions in this case.
Since -1 < sinx < 1 for all x, there are no solutions to sinx = \.
51
If sinx = then x = y or x = y5. For x = y, we have
7n 7n _ / V5\ _ 1 5V3 7V3

8 cos x + 5 tan x = 8 cos — + 5 tan — = 8 • —— + 5 • —=. -4V3
J
-
6 6 I 2 V3
For x = y11, we have
0 n 117T ll7I V3 ( 1 5V3 7V3

8 cos x + 5 tan x = 8 cos —— + 5 tan —— - 8 • —- + 5 --= = 4 V3 -
6 6 2 V V3
Therefore, the points of intersection are (x, y) =
3.62 Since cos2 x + sin2 x = 1, we have 1 - cos2 x = sin2 x, so
1 - sin4 x - cos2 x = sin2 x - sin4 x = sin2 x(l - sin2 x) = sin2 x cos2 x.
By the double angle formula for sine, we have sin 2x - 2 sin x cos x, so sin2 x cos2 x = sm^2*.
Hence, the given equation becomes su^— = ^, so sin2 2x = |, which means sin2x = ±\.
We are given that < x < |, so -n < 2x < n. The only angles in this range whose sines are ±\ are —
- _5n rt_ n_ anri 5n
and Therefore, the solutions are 17/ 17/ ‘try f O.LWX
12
3.63 From the sum-to-product formula for sine and the difference-to-product formula for cosine, we have
. 2x + y _ f
sin x + sin(x + y) = 2 sin y\ _ o2 sin 2x + y y
^ COS . ^ ^ cos 7-,
2 V 2
H)- 2 2
and
;„2x + y„;„( y\ n-lx + yy
cos x - cos(x + y) = -2 sin r, sin * 2 sin „ sin
2 V 2 2 2
so
2* + y y
sin x + sin(x + y)
2 sin cos
f(x) =
2
cos x - cos(x + y) 2^+ y
2 sin sm y
2
Thus, / is constant.
3.64 The given equation reminds us of the identity 1 + tan2 x = sec2 x. Rearranging this gives sec2 x - tan2 x = 1,
from which we have (sec x - tan x)(sec x + tan x) = 1. We are given sec x + tan x = y, so we have sec x - tan x = y.
Adding this to the given equation gives secx = ||j, and subtracting it from the given equation yields tanx = y|.
Therefore, we have cosx = || and sinx = cosxtanx = |||. This gives us cscx = ||| and cotx = = f§|, so
cscx + cotx =
R41
= 29
15
Challenge: see if you can find a less computational solution starting from the fact that (yyy) = 1.
3.65 The expression in x and y vaguely resembles some of our trig identities, particularly identities involving
tangent. (We think of tangent instead of sine or cosine because x and y can be any real number.) So, we let
52
Challenge Problems
x = tan A and y = tanB. Then
(x + y)(l - xy) (tan A + tanB)(l - tan A tan B)

(1 + x2)(l + yl) (1 + tan2 A)(l + tan2 B)
/ sin A . sinB \
_ V cos A cos B J
(1 - tan A tanB)
sec2A sec2 B
/ sin A sinB sin A sinB
(cos2 A cos2 B) 1-
V cos A cos B cos A cos B
/ sin A sinB sin A sinB
cos A cos B • (1 cosA cosB
V cos A cos B cos A cos B
= (sin A cos B + sinBcosA)(cosAcosB - sin A sinB)
= sin(A + B) cos(A + B)
= \ sin[2(A + B)],
which is always between -1/2 and 1/2.
3.66 Since 0 < 9 < we have 0 < sin 6 < 1 and 0 < cos 0 < 1. Then sin5 0 < sin2 0 and cos5 0 < cos2 0, so
sin5 0 + cos5 0 < sin2 0 + cos2 0 = 1.
But sin5 0 + cos5 0 = 1, so the inequalities above must be equalities. That is, sin5 0 = sin2 0 and cos5 0 = cos2 0.
Moving all the terms to one side in the first equation, we get sin5 0 - sin2 0 = 0, which factors as sin2 0 (sin3 0 -1) = 0.
Therefore, sin 0 = 0 or sin 0 = 1, which means 0 = 0 or ~ . Checking, we find that both of these solutions work.
Rearranging and factoring cos5 0 = cos2 0 leads to cos2 0(cos3 0 - 1) = 0, which leads to the same values of 0.
3.67 We can rewrite the given equation as tan 7x + cot 7x = sin 6x + cos4x. Let t = tan7x. Then cot7x = j, so
tan 7x + cot 7x = t + y. If t > 0, then by the AM-GM inequality, we have
t+ - > 2
t t'i = 2.
with equality if and only if t = 1. However, sin6x < 1 and cos4x < 1, so sin6x + cos4x < 2. Therefore, we must
have tan 7x = sin6x = cos4x = 1.
If t < 0, then let s = —f, so t + j = - (s + . Again by the AM-GM inequality, s + ^>2, so t + j < -2, with
equality if and only if t = -1. However, sin6x > -1 and cos4x > -1, so sin6x + cos4x > -2. Therefore, we must
have tan7x = sin6x = cos4x = -1.
Case 1: tan7x = sin6x = cos4x = 1.
Since 0 < x < 2n, we have 0 < 4x < 8n. The only angles in the interval [0, 8tz] whose cosines equal 1 are 0, 2n,
471, 6n, and 871, so x is equal to 0, f, n, or 2n. None of these values satisfies tan 7x = 1.
Case 2: tan7x = sin 6x = cos4x = -1.
The only angles between 0 and 871 whose cosine is -1 are n, 3n, 5n, and In, so x is equal to or
The only angles that satisfy both tan 7x = -1 and sin6x = -1 are f and
bn
Therefore, the solutions are x = and 4
3.68 First, we derive an angle sum identity for the cotangent function. From the angle sum identity for tangent,
1 1 - tan x tan y
cot(x + y) =-7-r = ----•
tan(x + y) tan x + tan y
53
To obtain a formula in terms of cot x and cot y, we divide both the numerator and denominator by tan x tan y:
1 - tan x tan y _ tan x tan y cot x cot y - 1

tan x + tan y 1 1 cot x + cot y
tan y tan x
cot x cot y - 1
Hence, cot(x + y) =
cotx + coty
Let a = arccot 3 and j6 = arccot 7. Then cot(« + j8) = = yyy = 2.
Let y = arccot 13 and 5 - arccot 21. Then cot(y + 6) = =
Then
10 cot(arccot 3 + arccot 7 + arccot 13 + arccot 21) = 10 cot[(n + j3) + (y + 6)]

_ 1Q cot(a + ft)cot(y + 5) - 1 _ 1Q 2-8-1
cot(g + jS) + cot(y + 6) 2+8
3.69 Let P - cos y cos — cos y. We note that in the sequence of angles f, y, and y, each angle is double the
previous angle. We can take advantage of this observation by using the double angle formula for sine. Multiplying
both sides by sin y, we get
. ti . n tl 2tl An
P sm — = sin — cos — cos — cos —.
7 7 7 7 7
By double angle formula for sine, we have 2 sin y cos y ~ sin so Psin y = \ sin y1 cos y1 cos y1. Similarly,
2 sin y1 cos y1 = sin y1, so P sin y = \ sin y1 cos y1. And again, 2 sin y1 cos y1 = sin y1, so P sin y = | sin yL But
y1 = 7i + y, and sin(7i + 0) = - sin 0 for all 0, so sin y1 = - sin y. Therefore, P = — g .
3.70
(a) By the triple angle formulas for sine and cosine, we have sin 3x = 3 sin x - 4 sin3 x and cos 3x = 4 cos3 x -
3 cos x, so
sin3x 3 sin x - 4 sin3 x sin x(3 - 4 sin2 x) 3-4 sin2 x
tan 3x = -—-^-— = tanx • --■=--.
cos 3x 4 cos3 x - 3 cos x cos x(4 cos1 x - 3) 4 cos^ x - 3
If we only had sin2 x and cos2 x terms in the numerator and denominator of the fraction, we could turn
them into constants and tan2 x terms by dividing the numerator and denominator by cos2 x. Therefore, we
write 3 as 3(sin2 x + cos2 x):
3-4 sin2 x _ 3(sin2 x + cos2 x) - 4 sin2 x _ 3 cos2 x - sin2 x

4 cos2 x - 3 4 cos2 x - 3(sin2 x + cos2 x) cos2 x - 3 sin2 x
We can then divide the top and bottom by cos2 x, to get
3 cos2 x - sin2 x _ 3 - _ 3 - tan2 x

cos2 x - 3 sin2 x 1 - 3 sin,2 * 1-3 tan2 x
COSz X
3 - tan2 x 3 tan x - tan3 x

Therefore, tan 3x = tan x •
1-3 tan2 x 1-3 tan2 x
54
Challenge Problems
2 _
(b) First, we determine sin 0. We have sin2 6 = 1 - cos2 0 = 1- (-A) = 1 ~ H = §§, so sin 0 = ± ygf = ± j§.
Since 0 lies in the second quadrant, we have sin 0 =
By the half-angle formula for tangent, we have
n0_ sin 0 _ 15/17 15 5

an 2 ~ 1 + cos0 ~ 1-8/17 “ ~9 ” 3'
Then by the triple angle formula for tangent, we have
30 _ 3tan(0/2) - tan3(0/2) _ 3 • (5/3) - (5/3)3 _5_

an 2 “ 1 - 3tan2(0/2) _ 1 - 3(5/3)2 99
(c) The angles 40° and 80° are both 20° from 60°, so we know that we can express tan 40° and tan 80° in terms
of tan 20°. So, we let t = tan 20°. Then by the angle sum identity for tangent, we have
tan 60° + tan 20° V3 + t

tan 80° = tan(60° +20°) =
1 - tan 60° tan 20° l-fV3'
and by the angle difference identity for tangent, we have
tan 60° - tan 20° V3 -t

tan 40° = tan(60° - 20°) =
1 + tan 60° tan 20° 1 + fV5'
so
t V3 + t a/3 - t K3 - t2)
tan 20° tan 40° tan 80°
l-fV3 l + fV3 1-312'
This result resembles our earlier triple angle identity for tangent, and 60° = 3 • 20°. So, we apply the triple
angle identity and find:
3 tan 20° - tan3 20° 3t -13 t(3 - tz)

tan 60° = tan(3 • 20°) =
1-3 tan2 20° 1-312 1-312'
Hence, tan x = tan 60°, so x = 60° .
3.71 Expressing all the trigonometric functions in terms of sin 0 and cos 0, we find
„ „ „ _ . _ ^ sin 0 cos 0 1 1
sin 0 + cos 0 + tan 0 + cot 0 + sec 0 + esc 0 - sm 0 + cos 0 +-- + ——- +-- +
cos 0 sin 0 cos 0 sin 0
sin2 0 + cos2 0 + sin 0 + cos 0
= sin 0 + cos 0 +
cos 0 sin 0
1 + sin 0 + cos 0
= sin 0 + cos 0 +
cos 0 sin 0
Let s = sin 0 + cos 0 and p = sin 0 cos 0. Then from the given equation, we have s + ^ = 7. Solving for s, we
find s = ■ T° solve for p and s, we must find another equation involving s and p.
Note that
s2 = (sin 0 + cos 0)2 = sin2 0 + 2 sin 0 cos 0 + cos2 0 = 1 + 2p.
Substituting s = ^ into this, we get ( j~[ ) = 1 + 2p, which simplifies as 2p3 - 44p2 + 18p = 0. This factors as
2p(p2 - 22p + 9) = 0. But p t 0 (because both sec 0 and esc 0 are defined), so p2 - 22p + 9 = 0. By the quadratic
formula, we have
22±V222 - 4-9 _22±8V7_llilV?
55
Since -1 < sin 9 < 1 and -1 < cos 9 < 1, we have p - sin 9 cos 9 < 1. But 11 + 4 V7 > 1, so p = 11 - 4 y/7.
Finally, by the double angle formula for sine, we have sin 29-2 sin 9 cos 9 = 2p = 22-8 V7
3.72 The right side of the recursion looks a lot like the angle sum identity for tangent. So, we let an = tan 9n, and
we note that
un + cin+1 tan 9n + tan 9n+\
tan(0n+2) — &n+2 = tan(0„ + 9n+1).
1 UnUn+i — 1 - tan 9„ tan 9n+\
Therefore, if we let 9n+2 = 9n+1 + 0,„ the recursion for an is satisfied when we let an - tan 9n. Flence, we can find
02009 by finding 02OO9-
Since a\ = 1 and a2 = we can let 9\ - 45° and 02 = 30°. We have 9\ = 45° = 3 • 15° and 02 = 30° = 2 • 15°,
so each 9n is an integer multiple of 15°. Let 9n = tn • 15°, so t\ - 3, f2 = 2, and f„+2 = f„+i + tn for all n > 0. Since
180° = 12 • 15°, and the tangent function has period 180°, it suffices to compute f2009 modulo 12. (That is, we only
have to find what the remainder is when we divide f2009 by 12.)
Computing the terms of the sequence {tn} modulo 12, we get
3,2,5,7,0,7,7,2,9,11,8,7,3,10,1,11,0,11,11,10,9,7,4,11,3,2,5,7,0,..
Once we see 3 immediately followed by 2 a second time, we know that the sequence is periodic, since each term is
determined by the two terms before it. Hence, the sequence {tn} modulo 12 has period 24 (that is, it repeats every
24 terms) and we have f2oo9 = tn = 0 (mod 12), so a2oo9 - tan(0 • 15°) = tan0° = [o].
3.73 Let 45 < k < 133. Then

sin 1° _ sin[(k + 1)° - k°]
sin k° sin(k + 1)° sin k° sin(k + 1)° ’
By the angle difference identity for sine, we have
sin[(k + 1)° - jfc°] = sin(fc + 1)° cos k° - cos(k + 1)° sin k°,
so
sin[(k + 1)° - k°] _ sin(k + 1)° cos k° - cos(k + 1)° sin k°
sin k° sin(k + 1)° sin k° sin(k + 1)°
sin(k + 1)° cos k° cos(k + 1)° sin k°
sin k° sin(k + 1)° sin k° sin(k + 1)°
_ cos k° cos(fc + l)°
sink0 sin(k + 1)°
= cotk° - cot(k + 1)°.
Summing over k = 45, 47, 49,..., 133, we get
sin T sinlc sinl°

-+- -+•••+- (cot45°-cot460)+(cot470-cot48°)H-h(cot 133°-cot 134°).
sin 45° sin 46° sin 47° sin 48° sin 133° sin 134°
But cot 46° = - cot 134°, cot 47° = - cot 133°,..., cot 89° = - cot 91°, cot 90° = 0, and cot 45° = 1, so each cotangent
on the right above cancels with another, except for cot 45°. Therefore, we have
(cot45° - cot46°) + (cot 47° - cot48°) + • • • + (cot 133° - cot 134°) = 1,
which means
sin lc sin 1° sinT
+ + = 1.
sin 45° sin 46° sin 47° sin 48° sin 133° sin 134°
56
Challenge Problems
Therefore,
1 1 1
+ + +
sin 45° sin 46° sin 47° sin 48° sin 133° sin 134° sinl°'
and the answer is n
3.74 Note that 8 cos3 4x cos3 x = (2 cos 4x cos x)3. By the product-to-sum formula, cos 4x cos x = (cos 3x+cos 5x)/2,
so (2 cos 4x cos x)3 = (cos 3x + cos 5x)3. Hence, the given equation becomes
cos3 3x + cos3 5x = (cos 3x + cos 5x)3.
Let a = cos 3x and b = cos 5x. Then
a3 + b3 = (a + by <=> a3 + b3 = a3 + 3a2b + 3ab2 + b3 <=> 3a2b + 3ab2 = 0 <=> 3ab(a + b) - 0.
Hence, cos 3x = 0, cos 5x = 0, or cos 3x + cos 5x = 0. But cos 3x + cos 5x = 2 cos x cos 4x, so cos 3x + cos 5x = 0 gives
cosx = 0 or cos4x = 0. So to summarize, x must satisfy one of the equations cosx = 0, cos3x = 0, cos4x = 0, or
cos 5x = 0.
If cosx = 0, then x = 360°?c + 90° or x = 360°k + 270° for some integer k. There are no solutions such that
100° < x < 200°. (Note that we can express “x - 360°fc+90°orx = 360°k+270° for some integer k" as "x - 180°fc+90°
for some integer k.")
If cos 3x = 0, then 3x = 360°k + 90° or 3x = 360°k + 270° for some integer k, sox = 120 °k + 30° or x = 120 °k + 90°.
The only solution such that 100° < x < 200° is 120° + 30° = 150°.
If cos4x = 0, then 4x = 360°k + 90° or 4x = 360°k + 270° for some integer k, so x = 90°k +|°orx = 90°k + ^ .
The only solutions such that 100° < x < 200° are 90° + f° = j° and 90° + - 3y° ■
If cos 5x = 0, then 5x = 360°k + 90° or 5x = 360°k + 270° for some integer k, so x = 72°k + 18° or x = 72°k + 54°.
The only solutions such that 100° < x < 200° are 72° • 2 + 18° = 162°, 72° + 54° = 126°, and 72° • 2 + 54° - 198°.
Therefore, the sum of the solutions such that 100° < x < 200° is
225° 315°
150° + ==^ + - + 162° + 126° + 198° 906°
2 2
3.75 Note that sin(90° - x) = cos x and cos(90° - x) = sin x, so
sin2 (90° - x) = cos2 x,

cos2 (90° - x) = sin2 x,
sin(90° - x) cos(90° - x) = cos x sin x.
Hence, sin2 (90° - x), cos2 (90° - x), and sin(90° - x)cos(90° - x) do not form the sides of a triangle if and only if
sin2 x, cos2x, and sinxcosx do not form the sides of a triangle. Thus, the answer is the same if we restrict our
attention to those values of x for which 0° < x < 45°.
For x such that 0° < x < 45°, we have cos x > sin x > 0, so cos2 x > sin x cos x > sin2 x. Hence, sin2 x, cos2 x,
and sin x cos x do not form the sides of a triangle if and only if sin x cos x + sin2 x < cos2 x, which is equivalent to
sin x cos x < cos2 x - sin2 x.
By the double angle formulas for sine and cosine, we have sin 2x = 2 sin x cos x and cos 2x = cos2 x - sin2 x,
so the inequality sinxcosx < cos2x - sin2x is equivalent to - cos2*/ which means < 2, or tan2x < 2.
Hence, 2x < arctan2 (since tan 6 is strictly increasing for 0 < 9 < 45°), so
arctan 2
57
Therefore, the probability that sin2 x, cos2 x, and sin x cos x do not form the sides of a triangle is given by
arctan2
_2_ arctan 2
45 90
3.76 The expressions in that terrifying equation we must prove strongly resemble the double angle formula for
tangent. So, we let x = tana, y = tanb, and z = tanc. From the double angle formula for tangent, we have
2 tan a 2x
tan 2a
1 - tan2 a 1-x2'
2 tan b 2y
tan lb
1 - tan2 b i -y2'
2 tanc 2z
tan 2c
1 - tan2 c 1-z2'
so we must show that

tan 2a + tan 2b + tan 2c = (tan 2a)(tan 2b) (tan 2c).
We must somehow use the given equation x + y + z - xyz, which we can write as tan a + tan b + tan c =
(tan a)(tan b)(tan c). We get expressions like tan a + tan b and (tan a)(tan b) from the identity for the tangent of a sum
of two angles. So, we wonder if we get expressions like tana + tan b + tanc and (tana)(tanb)(tanc) in an identity
for the tangent of a sum of three angles.
By applying the angle sum identity for tangent twice, we find
tan(a + b) + tan c
tan(a + b + c) tan[(a + b) + c]
1 - tan(a + b) tan c
tan a + tan b
1 - tan a tan b + tan C _ tan a + tan b + tan c - tan a tan b tan c
tan a + tan b 1 - (tan a tan b + tan a tan c + tan b tan c)
1 - tan a tan b
Aha! Since we are given tana+tanb+tanc = (tana)(tan b)(tanc), we now know that tan(a + b+c) = 0. Therefore,
a + b + c = nk for some integer k. Then la + lb + 2c = Ink, so
tan(2a + 2b + 2c) = 0.
But from our formula above for the tangent of the sum of three angles, we have
„ . tan 2a + tan 2b + tan 2c - tan 2a tan 2b tan 2c

tan(2a + 2 b + 2c) = -— -----------—.
1 — (tan 2a tan 2b + tan 2a tan 2c + tan 2b tan 2c)
Since this must equal 0, we must have
tan 2 a + tan 2b + tan 2c = tan 2a tan 2b tan 2c,
so we have
2x 2y 2z _ 2x 2y 2z
1 - x2 + 1 - y2 + 1 - z2 1 - x2 1 - y2 1 - z2 ’
3.77 First, to deal with the 4" term, we let bn = a„/2", so a„ = 2nb„. Substituting into the given formula, we get
= | • 2"bn + | n/r - 4»fc2 = | • 2”fc„ + 5.2”
Dividing both sides by 2”+1, we find
58
Challenge Problems
The term yjl-b* strongly suggests a trigonometric substitution, as do the | and |, since the sum of the squares
of these is 1. Since each an is real, each bn must be real. Therefore, we must have 1 - b2n > 0, so -1 < bn < 1, which
means we can let bn = sin an to take advantage of the identity sin2 an + cos2 an = 1. Letting bn = sin an therefore
gives us
bn+l
4 sin
- • an + -3 yA-4 • an + -3 \f/-
1 - sin an = - sin cos17- 4 sm
an = - • an + -1
3I cos an
5 5 5 5 v 5 5
We can take further advantage of trigonometric identities by noting that because (|) +(|) = A there is an angle
0 such that sin 0 = | and cos 9 = |. Therefore, we can write the equation above as
sinn„+i = cos0sina„ + sin0|cosa„|.
By the angle sum and angle difference identities for sine, we have
sin(a ± ft) = sin a cos jS ± sin j8 cos a,
so
f sin(0 + an) if cosn„ > 0,
\ sin(0 - an) if cos an < 0.
We have bo = sin0°, so we can take ao = 0°. Then cos «o - cos 0° = 1 > 0, so b\ = sin 0. We can take a\ = 0.
Then cos cc\ = cos 0 = | > 0, so bi = sin 20. We can take «2 - 20.
To find b3/ we must find the sign of cos 20. By the double angle formula for cosine, we have cos 20 =
cos2 0 - sin2 0 = ^ > 0, so b3 = sin 30. We can take a3 = 30.
To find b4, we must find the sign of cos 30. By the triple angle formula for cosine,
/4\ 3 4 44
cos30 = 4cos3 0 - 3cos 0 = 4- ( -J -3• - = < 0/
so &4 = sin(30 — 9) — sin 20.
But since bn+x depends only on the value of bn, and b4 = sin 20 = b2, the sequence {bn} is periodic starting with
the term b4, with period 2. Therefore, by the double angle formula for sine, we have
3 4 24
bw = b2 = sin 20 = 2 sin 0 cos 0 = 2
5 ' 5 25'
24576
SO U\o = 210bio = 25
3.78 First, we derive an angle sum identity for arctangent. Let a and be acute angles, and let x - tan a and
y = tan /3, so x and y are positive. Then by the angle sum identity for tangent, we have
tan a + tan j8 x+y

tan(a + ft)
1 - tan a tan f> 1 -xy'
If xy < 1, then (x + y)/( 1 - xy) is positive. Furthermore, 0 < a < | and 0</3<f, so 0 < a + fl < n. Since
tan(n + f) = (x + y)/(l - xy) is positive, a + ^ must be an acute angle. Therefore, we may take the arctan of both
sides, to get
x+y
a + fl = arctan
1 -xy'
Since a = arctan x and /3 = arctan y, the equation becomes
x+y
arctan x + arctan y = arctan --.
17 1 - xy
59
This formula holds for all positive real numbers x and y such that xy < 1.
Let Sn = 2_j arctan —. We will prove that Sn - arctan ■ for all positive integers n by induction.
k=1 n +
For n = 1, we have Si = arctan(l/2), so the result holds for the base case n = 1. Assume that the result holds
for some positive integer n = m, so
_ A 1 m
Sm = 2_^ arctan —^ = arctan
m+ 1’
k=1
Then
r ^ 1 A 1 1 m 1
Sm+i = > arctan —z = > arctan —r + arctan —-—r- = arctan-- + arctan —- 2
2/c2 ^ 2k2 2(m +1)2 m+ 1 2 (m +1)
2(m 4 , ‘
fc=i Jt=i
Both m/(m + 1) and l/[2(m + l)2] are positive, and
m 1 m
< 1/
m+1 2 (m + l)2 2(m + l)3
since m < 2(m + l)3. Therefore, by our angle sum identity for arctangent, we have
c m 1 L + 2(^W l (m + l)[2m(m +1) +1]

SOT+i = arctan-- + arctan —-= arctan i --—-—i—
_ m _ l = arctan
m+1 2 (m + l)2 1 m+1 ‘ 2(m+l)2
2(m + l)3 - m
(m + l)(2m2 + 2m + 1) v._ + l)(2m2 + 2m + 1)

(m (m + l)(2m2 + 2m + 1)
- arctan ——r—-—-— -—-= arctan ——=—-—-— -— = arctan
2(m3 + 3m2 + 3m + 1) - m 2m3 + 6m2 + 5m + 2 (m + 2)(2m2 + 2m + 1)
m+1
= arctan
m + 2'
Hence, the result holds for n = m + 1, and by induction, it holds for all positive integers n.
60
CHAPTER 4 i
l Applications to Geometry
4.1.1 We have sin ZPRQ = p^, so PR = ^PQ 8 PQ. 8

sin /.PRQ sin 42° 11.96 , and tan iPRQ = so QR = tan ZPRQ tan 42°
8.88
4.1.2 We have tan LXZY - so ZXZY = arctan YZ

= arctan4 » 76°
4.1.3 The line 2x + 3y = 7 has x-intercept (7/2,0) and y-intercept (0,7/3). Therefore,
if 6 is the acute angle formed by the x-axis and the line 2x + 3y = 7, then
7/3 _ 2
tan 9 =
7/2 ~ 3'
so 6 - arctan(2/3) ~ 34°
We could also have tackled this problem by using the relationship between tan¬
gent and slope that we proved in the text. Letting <p be an angle between the line and the x-axis, we have
tan cp = -2/3, from which we find <p = arctan(-2/3) » -34°, which tells us that the acute angle between the x-axis
and the line is approximately 34°.
4.1.4 We see that BC = 12/4 = 3, so the area of triangle ABC is \ AB ■ BC sin lABC = \ • 12 • 3 • \l _
9
4.1.5 First, we calculate the horizontal distance between the two peaks (that is, the 6336 feet
distance as viewed from above, without the vertical component). On the map, the
two peaks are 1 inch apart, so in real life, the horizontal distance between them is
1.2 miles 5280 feet c x

1 inch • ———-— • ——-— = 6336 feet.
1 mch 1 mile
Then the vertical distance between the two peaks (in feet) is 6336 • tan 4° ~ 440. Therefore, the second peak is
14000 - 440 13560 feet above sea level
4.1.6 Since AO = 1 and LOCA is right, we immediately have CO = cos 9 and AC - sin 9 \. Since AO = 1 and
lOAD is right, we have AD/AO - tan 9, so AD - tan 9 [ and AO/DO = cos 9, so DO = sec 9
Likewise, lOAB is right. Also, lAOB = iBOD - /.AOD = 90° — 9, so ZABO = 90° - ZAOB = 9. Hence,
AO/AB = tan 9 (and we still have AO = 1), so AB = cot 9 , and AO/BO = sin 9, so BO -- esc 9
61
CHAPTER 4. APPLICATIONS TO GEOMETRY
4.1.7 Let the side length of the cube be s. Since AC is the diagonal of a square with side
length s, we have AC = s V2. Then by the Pythagorean Theorem on right triangle ACB, we
have AB = VAC^TBC2 = V2s2 + s2 = V3s2 = s V3.
Therefore,
BC s 1 V3
cos /.ABC =
AB sV3 V5 3
4.1.8 Let M be the midpoint of AB, so OM = AB/2 = 450. Let x = EM and y = EM, D C
so x + y = EM + EM = EF = 400.
Let n - zEOM and jS = ZfOM. Then « + /? = ZEOM + ZEOM = ZEOE = 45°,

so tan(n + fi) - tan 45° = 1. But by the angle sum identity for tangent, we have
tan(a + jS) = Since ZEMO and ZEMO are right, we have tann = ^
and tan ^ . Therefore, we have
tan (ft + g) = tann + tan/j = iso + 450 = 45(X* + y) = 180000

1 - tan a tan jS 1 “ ifo ' 450 202500 - xy 202500 - xy'
where we used x + y = 400 in the final step. We also have tan(« + /?) = 1, so
180000 = 202500 - xy, which implies xy = 22500. Substituting x = 400 - y from earlier into xy = 22500 gives
(400 — y)y = 22500. Rearranging this equation gives y2 - 400y + 22500 = 0. By the quadratic formula, we have
_ 400 ± V4002 - 4 • 22500 _ 400 ± V70000 _ 400 ± 100 V7 _ OQ0 ( 5Q c
Since AE < BE, we have x > y, so x = 200 + 50 V7 and y = 200 - 50 V7. Therefore, we have
BE = EM - ME = 450 - (200 - 50 V7) 250 + 50 V7
4.2.1 Let triangle ABC be obtuse, where ZC is obtuse. Let X be the foot of the B
altitude from B to AC.
We see that ZBCX = 180° - /BCA = 180° - ZC, so CX = BC cos ZBCX = a cos(180° -
ZC) = -acosC, and BX = BC sin ZBCX = asin(180° - C) = asinC. (Note that since
ZC is obtuse, cos C is negative, so the expression CX = -a cos C makes sense.)
By the Pythagorean Theorem on right triangle BXA, we have AB2 = BX2 + AX2. Since AB = c, BX = asinC,
and AX = AC + CX - b - a cos C, we have
c2 = (a sin C)2 + (b - a cos C)2 = a2 sin2 C + b2 - lab cos C + a2 cos2 C

= (a2 sin2 C + a2 cos2 C) + b2 - lab cos C - a2 + b2 - lab cos C.
Thus, the Law of Cosines holds for an obtuse triangle. The proof in the text suffices when ZB is obtuse or right. If
ZA is obtuse or right, we can also use the proof in the text by simply swapping a and b, and swapping A and B,
throughout the proof given in the text.
62
Section 4.2
Next, let triangle ABC be right, where ZC is right. Then by the Pythagorean Theorem, AB2 = BC2 + AC2, or
c2 = a2 + b2. But cos C = cos 90° = 0, so the equation c2 = a2 + b2 - lab cos C holds. Thus, the Law of Cosines also
holds for a right triangle.
4.2.2 Let A be the airport, W be the location of the westward-flying plane 90 minutes after takeoff, and E be
the location of the eastward-flying plane 90 minutes after takeoff. Since the westward plane flies due west and
the eastward plane flies 40° north of east, we have IWAE = 90° + (90° - 40°) = 140°. The westward plane flies
at 200 miles per hour for 90 minutes, which is 1.5 hours, so WA = 1.5(200) = 300 miles. Similarly, we have
EA = 250 • 1.5 = 375 miles. Applying the Law of Cosines, we have
WE2 = EA2 + WA2 - 2(WA)(EA) cos IWAE « 403,000,
so WE 635 miles
4.2.3
(a) By the Pythagorean Theorem, AB2 = BC2 + CA2 if and only if lACB is right.
(b) By the Law of Cosines, cos lACB = ■ Therefore, AB2 > BC2 + CA2 if and only if cos lACB < 0.
But lACB is an angle of a triangle, so cos lACB < 0 if and only if lACB is obtuse.
(c) By the Law of Cosines, AB2 < BC2 + CA2 if and only if cos lACB > 0. But lACB is an angle of a triangle, so
cos lACB > 0 if and only if lACB is acute.
4.2.4 Let the lengths of the sides be x, lx, and 6.
If the angle opposite the side of length x is 120°, then by the Law of Cosines,
(lx)2 + 62 - 2 • 2x • 6 cos 120° = x2 => 4x2 + 36 + 12x = x2 => 3x2 + 12x + 36 = 0.
This equation clearly has no positive solutions in x.
If the angle opposite the side of length 2x is 120°, then
x2 + 62 - 2 • x • 6 cos 120° - (2x)2 => x2 + 36 + 6x = 4x2 => 3x2 - 6x - 36 = 0 => x2 - 2x - 12 = 0.
By the quadratic formula, x = 2±V2^+4'12 = = 2±2^> = 1 ± Vl3. Since x must be positive, the only solution in
this case is x = 1 + Vl3.
If the angle opposite the side of length 6 is 120°, then
36
x2 + (2x)2 - 2 • x • 2x cos 120° = 62 => x2 + 4x2 + 2x2 = 36 =» 7x2 = 36 => x2 = —,
so x = 6/ V7 = (6 V7)/7.
Therefore, the possible side lengths of the triangle are 1 + Vl3, 2 + 2 Vl3, and 6 and 6 V7/7,12 V7/7, and 6
4.2.5 By the Law of Cosines, cos iXYZ = XY ^xy-yz^2 = 3 2%57 ~ TkF = ~\> so 2XYZ = 120 °
4.2.6 Let 6 - lABC. Then by the Law of Cosines on triangle ABC,
AC2 = AB2 + BC2 - 1AB ■ BC cos 6 = 22 + 32 - 2 • 2 • 3 cos 6 = 13 - 12 cos 6.
Since quadrilateral ABCD is inscribed in a circle, we have iCDA = 180° - lABC = 180° - 6. Therefore, by the
Law of Cosines on triangle ADC, we have
AC2 = CD2 + DA2 - 1CD ■ DA cos(180° - 0) = 42 + 62 - 2 • 4 • 6(- cos 6) = 51 + 48 cos 6.
63
Hence, we have
„ 39 13
13 - 12 cos 6 = 52 + 48 cos Q => 60 cos 6- -39 =» cosd--—
6U 20 '
2V130
Then AC2 = 13-12 cos 0 = 13 + 12-^ = ^ = so AC = = 5
4.2.7 Let OP denote the circle with center P for any point P. Let OB, OC, and OD X
be tangent to OA at X, Y, and Z, respectively. Let r be the radius of OE.
Since the radius of OA is 10 and the radius of OC is 2, we have AC = 8 and

CE - r + 2. By symmetry, E lies on the diameter of OA passing through X. Since
the radius of OB is 3, we have AB = 7 and BE = r + 3, so AE - BE - AB = r - 4.
Triangle XYZ is equilateral and A is its center, so ZXAY = 120°, which means
ZCAE = 60°. Therefore, by the Law of Cosines on triangle CAE, we have
AC2 + AE2 - 2AC ■ AE cos 60° - CE2
=> 82 + (r - 4)2 - 2 • 8 • (r - 4) • ^ = (r + 2)2
, 108
=> 64 + r2 - 8r + 16 - 8r + 32 = r2 + 4r + 4 20r --108 => r = —— =
4.2.8 Let x - PC. Since lAPB, iBPC, and ZCPA are all equal and add A
up to 360°, each angle is equal to 120°. Then by the Law of Cosines on
triangles PBC, PAB, and PAC, we have
a2 = x2 + 62 - 2 • 6x cos 120° - x2 + 6x + 36,

c2 = 62 + 102 - 2 • 6 • 10 cos 120° - 196,
b2 - x2 + 102 - 2 • lOx cos 120° = x2 + lOx + 100.
However, by the Pythagorean Theorem, we have a2 + c2 = b2. Substituting the expressions above for a2, b2, and c2
gives
x2 + 6x + 36 + 196 = x2 + lOx + 100 => 4x = 132
4.3.1 Let A, B, and W denote my house, my neighbor's house, and the well, respectively. Then from the given
information, AW = 150 (in feet), ZBAW = 42°, and ZAWB = 123°. Then ZABW = 180° - ZBAW - ZAWB =
180° - 42° - 123° = 15°. Therefore, by the Law of Sines, we have
AB AW An AW sin ZAWB 150 sin 123° r——-

sin ZAWB sin ZABW sin lAB\N sin 15° 1
64
Section 4.3
4.3.2 Let A and B denote the top and base of the tower, let S denote the tip of the A
shadow, and let G denote the point on the ground directly below A. Then from the
given information, BS = 20, LABG = 85°, and LASB = 72°.
Then LABS = 180° - LABG = 180° - 85° = 95°, and LBAS = 180° - LASB - LABS =
180° - 72° - 95° = 13°, so by the Law of Sines on triangle ABS,
AB BS BS sin LASB 20 sin 72°

AB = 84.6.
sin LASB sin LBAS " sin LBAS sin 13°
Then from right triangle AGB, we have AG = AB sin LABG = 84.6 sin 85° 84 feet
4.3.3 We claim that AB > BC > CA if and only if zC > LB > LA. (In other words, the
converse is also true.)
First, we cover the case of an acute or right triangle. By the Extended Law of Sines, we have AB = 2R sin C,
AC = 2RsinB, and BC = 2RsinA, so AB > AC > BC if and only if sinC > sinB > sin A. But the sine function is
increasing on the interval [0, \ ], so sin C > sin B > sin A if and only if LC > LB > LA.
Next, we cover the case of an obtuse triangle. Assume that ZC is obtuse, so ZC is the largest angle of the
triangle. Furthermore, by the Law of Cosines,
AB2 = AC2 + BC2 - ZAC • BC cos C.
Since ZC is obtuse, cos C is negative, so -2AC • BC cos C is positive. Therefore, AB2 > AC2 and AB2 > BC2, which
implies AB > AC and AB > BC. Hence, AB is the largest side of the triangle.
Since ZC is obtuse, both LB and ZC are acute. Hence, by the same argument as in the acute triangle case,
AC > BC if and only if LB > LA. Therefore, for an obtuse triangle, AB > AC > BC if and only if LC> LB > LA.
4.3.4 From the given information, we have sB?A+sm_B+sinC _ a±Lfcc_ By the Extended Law of Sines, we have
a = 2R sin A, b - 2R sin B, and c = 2R sin C, so
sin A + sinB + sinC a+b+c 2RsinA + ZRsinB + 2RsinC R, . .

-3-= —12— =-T3-= r(sm A + sinB + smC),
so R = 2.
Suppose that a = 2. Then sin A = ^ = |, so ZA is 30° or 150°
4.3.5 Applying the Law of Sines gives sm^CB = suyic^~' so sin 4ACB = ^ sin LCAB = p sin 30° = We have
arcsin ^ w 20.9°, so the possible values of LACB are 20.9° or 180° - 20.9° = 159.1°. However, since LCAB = 30°,
we cannot possibly have LACB = 159.1°, since the sum of these two angles is greater than 180°, which we cannot
have in A ABC. So, there is only one possible value for LACB. That is, if AB = DE = 5, BC = EF - 7, and
LCAB = LFDE = 30°, then we can deduce that A ABC = A DEF. So, this wouldn't be an appropriate example to
show why there isn't an SSA Congruence Theorem.
4.3.6 We cannot use the Law of Sines to prove AA Similarity because our very use of sine assumes that AA
Similarity is true. For example, to show that every right triangle with a 23° angle has the same ratio
leg opposite the 23° angle

hypotenuse
we use AA Similarity. In other words, we use AA Similarity to show that sine "works." So, we can't turn around
and then use sine to show that AA Similarity works.
65
4.3.7 Solution 1. We have /CAP = /CBP = 10° and CA = CB (as radii of

the same circle). The Law of Sines applied to A CAP and A CPB gives us
sin ZCPA sin /.CAP sin /CPB sin /CBP

and
AC CP CB CP
Combining these with the equalities above, we have sin /CPA = sin /CPB.
However, triangles ACP and BCP are NOT congruent, and /CPA ± /CPB,
so sin /CPA = sin /CPB tells us that /CPA + /CPB = 180°.
Since /ACM = 40°, we have /ACP = 180° - /ACM = 140°. Then /CPA = 180° - /CAP - /ACP = 180° -
10° - 140° - 30°. Hence, /CPB = 180° - /CPA = 180° - 30° = 150°, so /BCN = /BCP = 180° - /CBP - /CPB =
180° - 10° - 150° = 20°. Therefore, we have BN = /BCN = 20 °
Solution 2. As in Solution 1, we have /CPA = 30°. By the Law of Sines on triangle CPA,
AC CP CP
sin /CPA sin /CAP ~ sin 10°
By the Law of Sines on triangle CPB,
BC CP CP
sin /CPB sin /CBP sin 10°
Since AC = BC, it follows that sin /CPB = sin /CPA = sin 30°. Since /CPB is obtuse, and sin 150° = sin 30°, we
conclude that /CPB = 150°. Then again as in Solution 1, we have BN = /BCN = 20 °
in length. Let x = AE = AF. Similarly, let y = BD = BF, and let z - CD = CE.
Hence,
x + y = AF + BF = AB = c,
x + z = AE + CE = AC = b,
y + z - BD + CD = BC = a.
Adding the first two equations, and subtracting the third equation, we get
(x + y) + (x + z) - (y + z) = b + c - a,
which simplifies as 2x - b + c - a. Hence, x = = a+b+^~2a = ^±|±£ - a = s - a.
Similarly, adding the first and third equations, and subtracting the second equation, we get
(x + y) + (y + z) - {x + z) = a + c - b,
_ a+c-b _ a+b+c-2b _ n+b+c
which simplifies as 2y - a + c - b. Hence, y -b = s-b.
66
Section 4.4
4.4.2 Let I be the incenter of triangle ABC. Then triangle IBC has height r and base A
BC = a, so
UBC] = \ra.
Similarly, [MC] = \rb and [MB] = \rc, so
[ABC] = [IBC] + [IAC] + [MB] = ^ra + ^rb + ~rc = r ■ — ^ + C - rs.
altitude from B to AC.
We see that zBCX = 180° - zBCA = 180° - ZC, so CX = BC cos zBCX = a cos(180° -
ZC) = —a cos C. Also, AX - AB cos LB AC - c cos A. Therefore, b = AC = AX - CX =
labels A and C, and swap a and c, throughout. The proof in the text suffices when ZB is obtuse. Thus, the formula
holds for an obtuse triangle.
Next, let triangle ABC be right, where ZC is right. Then b = AC = AB cos ABAC = ccos A. But cos C = cos 90° =
0, so b = c cos A + a cos C. Thus, the formula also holds for a right triangle. As with the obtuse case, this proof
addresses the case in which ZA is right, and the proof in the text covers the case in which ZB is right.
4.4.4 By the Law of Sines, we have sin A = sin B = and sin C = so
. . . _ . _ a b c abc
sin A sm B sin C =
2R 2R 2R 8 R3'
Since abc - 4R[ABC], we have sin A sin B sin C = = 4Rg^3C] - -^|p.
4.4.5
(a) The desired formula resembles the formula we already found for cos y, so we write sin j in terms of cos j
and use the formula cos j = that we proved in the text:
. ,A „ 7 A „ s(s-a) bc-s(s-a)
sin — = 1 - cos — = 1-7-=-t-.
2 2 be be
We now must show that the numerator of that fraction equals (s - b)(s - c). A little algebra does the trick:
be - s(s - a) = be - s2 + as,
(s - b)(s -c) = s2 ~(b + c)s + bc = s2 - (2s - a)s + bc -bc~s2 + as.
So, we have sin ^ = ± y/(s ^ c) . But 0° < A < 180°, so 0° < j < 90°, which means sin j is positive. Hence,
. A (s - b)(s - c)
sm - =
be
(b) By the double angle formula for sine, we have sin A = 2 sin j cos so
1 A =bc. /(s-frXs-c). /s(s-fl) = \Js(s — a)(s — b)(s — c).

[ABC] =-&c sin A =-fee-2 sin-cos-\ bc \ bc
67
(c) From the formula [ABC] = ||, we have
R_ a^c _ a^c _ fl^c __abc_

4[ABC] 4 Vs(s - fl)(s - b)(s - c) yJ16s(s - a)(s - b)(s - c) >J2s(2s - 2a)(2s - 2b)(2s - 2c)
abc
V(fl + b + c)(-a + b + c)(a - b + c)(« + b - c)
4.4.6 Let d = AM. Then by Stewart's Theorem, we have
( -S) Cl Cl \ Cl ,9 $ b +
2 b2 + a“•
(2 + - • - = X • b2 + - • c d2 + — = d2 =
V 2 2/22 2 J
2b2 + 2c2 - a2 'Jib2 + 2c2 - a2

d2 = d=
4.4.7 Let X, Y, and Z be the feet of the altitudes from A, B, and C, respectively.
We can express AH in terms of other lengths and trig functions of angles with right
triangle AYH, which gives us sin ZAHY - so AH - sin^JHy ■ Searching for other
angles that equal ZAHY, we note that
LAHY = 90° - lYAH = 90° - zCAX = 90° - (90° - C) = C,
B X C
so AH = -4A
sin L
Now we can use the Extended Law of Sines to introduce R. The Extended
Law of Sines applied to A ABC gives 2 R = so AH = We wish to show that AH = 2KcosA,
so all we have left is to show that cos A = This follows immediately from AABY, in which we have
cos A = cos ZYAB = Therefore, we have
(2 R)(AY) =2R'AY
AH = = 2Bcos A.
c c
4.4.8 We have
A-B
sin
A-B A-B
tan COS sin cos
A+B A+B
tan sin sin ^ cos ^
A+B
cos
Applying product-to-sum identities gives
A-B A+B 1 1
sm cos -[sinA + sin(-B)] = -(sinA - sinB),
2 2
. A+B A-B 1
sm —-— cos ——— -(sin A + sinB).
tan ^ sin ^ cos ^ sin A - sin B

Therefore -—
nererore, ^ ^ =---—
sin cos a^b —-
sin A + sin B ■
Finally, by the Extended Law of Sines, we have sin A = ^ and sin B = so
sin A - sinB _ ^ ^ _ a-b

sin A + sin B A + A a + b’
68
Review Problems
Review Problems
4.26 Since Zli = 90°, we have ZS + ZT = 90°. Combining this with the given ZS = 3ZT produces 4zT = 90°, so
ZT = 92- = 22.5°. Since cos ZT = we have
ST 1
1.08
TU cos 22.5°
4.27 The area of triangle PQR is given by
^PQ ■ QR sin IPQR = ^ • 12 • 4 V3 sin ZPQP = 24 V3 sin ZPQP.
We are given that the area of triangle PQR is 12 V6, so sin iPQR = ^. Therefore, the possible values of
ZPQK are 45° and 135°
4.28 Since IJLK - 90°, we have cos IJKL = j~ = so IJKL = arccos ^

12
54°
4.29
(a) Let a be the acute angle between k and the x-axis. The slope of t
line k is 2/3, so a - arctan | ~ 34°

(b) Let /> be the acute angle between ( and the x-axis. The slope of
k 'U
line i is 3, so = arctan 3 w 72°

(c) Let y be the acute angle between k and £, and consider the
triangle formed by k, l, and the x-axis. The angles of this triangle
are a, 180° - and y, so a + (180° - f}) + y = 180°. Therefore,
y = p - a « 72° - 34° « [38°
(d) Let a be the angle between the line with slope ni\ and the x-axis such that tana = m\. Similarly, let /I be the
angle between the line with slope mi and the x-axis such that tan = mi-
As in part (c), one of the angles between the two lines is jS - a. If we let y = j8 - a, then by the angle
difference formula for tangent, we have
tan j3 - tan a mi - m\
tan y = tan(/3 - a) =
1 + tan j8 tan a 1 + mi m2
We test the formula with lines k and €. The slope of k is 2/3 and the slope of l is 3, so taking m\ = 2/3 and
mi = 3, we get
3 - 2/3 7/3 7
tany
1 + 2/3 • 3 9'
soy = arctan(7/9) « 38°, as before.
4.30 By symmetry, the overlap region is a rhombus. Let s be a side length of the rhombus.
Then, we form a right triangle as shown. From this right triangle, we see that sin a =
so s = 1 / sin a = esc a. The area of the rhombus is simply the base times the height, which
is (l)(s) = esc a
69
4.31 As the boat sails along the latitude 30° N, it follows the arc of a circle on the
surface of Earth, so first we must find the length of that arc. Let P be denote the
center of this circle, let A be the position of the boat, and let O denote the center of
the Earth. We take a cross-section of the Earth that includes P, A, and O.
Since the boat is at 30° N latitude, we have lOAP = 30°. Since the radius of
the Earth is 3950 (in miles), we have PA = OA cos 30° = 3950 • ~ w 3420. The
circumference of circle O is then 2n • 3420 « 21500.
As the boat sails from 19° W to 46° W, it sweeps out an arc of 46° - 19° = 27°.
Therefore, the boat sails a distance of ■ 21500 ~ 1610 miles. Finally, the boat sails
at 9 miles per hour, so the travel time is 1610/9 - 179 hours
4.32 Let x = CM. Then BM = CM sin IMCB - 0.8x - y, and by the Pythagorean
Theorem, we have
3x
BC= Vca/pmbmp =
5~'
Since M is the midpoint of AB, we have AB = 2BM = —. Therefore, again by the Pythagorean Theorem,
8xV 3x V 64x2 9x2 73x2 :V73

AC = VaB2 + BC2 =
~5J + ~5J ~25” + 25" ~25~
3x
BC 5 3 V73
Hence, cos lACB =
AC xV73 V73 73
4.33 By the Pythagorean Theorem, we have AB - 5 and BD = 13. We have DE/DB = sin iDBE. Since
iDBE = 180° - LABC — lABD, we have
sin ZDBE - sin(180° - lABC - lABD) = sin(ZABC + ZABD).
By the angle sum identity for sine, we have
sin(ZABC + lABD) = sin lABC cos lABD + cos lABC sin lABD
AC AB BC AD_3 5^ 4 12 _ 15 48 _ 63
AB BD + AB'BD~5'l3 + 5'l3~65 + 65~ 65
4.34
(a) We can build a right triangle with iGIH as one of the acute angles by drawing altitude H
HM as shown. Because GH = HI, point M is the midpoint of GI, so GM = GI/2 = 3.
Then by the Pythagorean Theorem, we have HM = ^GH2 - GM2 = V72 = 6 V2, so
tan IGIH = HM/IM = 2V2

(b) We have
GM 1
tan IGHM =
HM 6V2 2V2'
Also, iGHI = 2lGHM, so by the double angle formula for tangent.
2 1
2 tan IGHM ■
z 2V2
—
Cl Cl __
8 8V2 4y/2
tan IGHI = tan(2IGHM) = 7
1 - tan2 IGHM 11 -18 7V2 ~ 14
8
Cl)
70
Review Problems
92 + 92 - 62 126
(c) By the Law of Cosines, we have cos iGHI
2-9-9 162
4.35 By the Law of Cosines, the distance between the two peaks is
V4.62 + 5.82 - 2 • 4.6 • 5.8 cos 37° 3.5 miles
4.36 In the text, we used the Pythagorean Theorem to prove the Law of Cosines. In the "proof" offered in this
problem, we then use the Law of Cosines to prove the Pythagorean Theorem. Because we used the Pythagorean
Theorem to prove the Law of Cosines, we cannot use the Law of Cosines to prove the Pythagorean Theorem. (If
we do, then we have essentially used the Law of Cosines to prove the Law of Cosines, which is clearly absurd!)
We say this proof is guilty of "circular reasoning," and is therefore not a valid proof.
4.37 Label the vertices of the triangle A, B, and C, so that AB = AC = 2 and BC = V6

- V2. Then by the Law of
Cosines,
. 22 + 22 -(V6-V2)2 4 + 4 - 6 + 4 V3 - 2 _ 4 V3 _ V3
Hence, A - 30°. Since AB = AC, we have ZB = ZC = (180° - 30°)/2 = 75°. Therefore, the angles of the triangle are
30°, 75°, and 75° .
4.38 Let* = AD. By the Law of Cosines on triangle ABD, we have cos A = = B
^. (We can also see that cos A = ^ by noting that the altitude from B to AD bisects
AD, because aABD is isosceles with AB - BD. So, we have cos A = = ^.)
By the Law of Cosines on triangle ABC, we have cos A = 5 \9^g7- = so

x/10 - 19/30, which means x = 19/3. Then CD = 9 - x = 8/3, so AD/DC =
(19/3)/(8/3) = 19/8
4.39 Let the side lengths of the triangle be * - 1, x, and x + 1. Since one angle is 60°, the other two angles must
add up to 120°, so let them be 60° - 9 and 60° + 9, where 9 is positive.
Since the angle with measure 60° is the "middle angle" (i.e., one other angle is larger and one is smaller), it
must be opposite the side of with the "middle length", which is x. So by the Law of Cosines, we have
l
{x - l)2 + (x + l)2 - 2(x - l)(x + 1) cos 60° = x2 x2 - 2x + 1 + x2 + 2x + 1 - 2(x2 - 1) - - = x2
=> 2x2 + 2- (x2 - 1) = x2 1 = 0,

which gives us a contradiction. That is, we have shown that if one angle of a triangle has measure 60° and the
three side lengths of a triangle form an arithmetic sequence with common difference 1, then we have 1 = 0. But it
is impossible to have 1 = 0, so we conclude that no such triangle exists.
4.40 Let ABCD be a regular tetrahedron of side length 1, and let M be the midpoint of
AB, so CM and DM are altitudes of equilateral triangles ABC and ABD, respectively. Then
the dihedral angle between faces ABC and ABD is equal to ZCMD, since CM and DB are
perpendicular to AB by symmetry.
Since AC ^ AD = 1, we have CM = DM = sin 60° = . Then by the Law of Cosines on

triangle CMD,
cos ZCMD =
f) +PP) -i2
a2 Vi
2
71
Therefore, iCMD = arccos(|) 70.5°
4.41 By the Law of Sines, we have S=&

QR
so Tin r
4
■A- = -A~.
sin P sin 2R
By the double angle formula for sine, we
J °
have sin 2R-2 sin R cos R, so
cos R =
sin R 2 sin R cos R 2-4
BC
4.42 By the Law of Sines, we must have sin A'
SO —
sin 57°
" isfe- However, ^ * 3.6 and
-AL
sin A
— —--8.4,
sin 73°
so no such triangle exists.
4.43 Let A, B, and C be the points where the plane starts, turns,
and ends, respectively. Then AB = 350 (in miles), AC = 800, and
ZB == 180° - 20° = 160°. Then by the Law of Sines,
AC AB AB sin B 350 sin 160°

sinC = 0.150,
sin B sin C AC 800
so ZC = arcsin 0.150 « 8.6°. Then ZA - 180° - LB - LC ^ 180° - 160° - 8.6° = 11.4°.
Again by the Law of Sines, we have
BC AC AC sin A 800 sin 11.4°

BC 462.
sin A sinB sinB sin 160°
The speed of the plane then is 462 miles/2 hours = 231 miles per hour, so the total flight time is
350 miles + 462 miles

3.5 hours
231 miles per hour
4.44 Label the vertices of the triangle A, B, and C, so that AB = AC = 12 and BC = 6. We want A
to find the circumradius of triangle ABC. From the Extended Law of Sines, the circumradius of
triangle ABC is given by R- 2^c- So, all we have to do is find sin C. We can build a right triangle
with ZC as one of the acute angles by drawing altitude AD. The foot of this altitude is the midpoint
of BC because A ABC is isosceles. The Pythagorean Theorem gives us AD = VAC2 - CD2 = 3 Vl5.
Therefore, we have sinC = yf = so B C
AB _ JL2_ 24 _ 24 Vl5 _ 8 Vl5

2 sin ZC VH Vl5 ~ 15 5
2
4.45 Solution 1. By the Law of Cosines, the equation AB cos B = AC cos C becomes
a „1
a2 +c2 -b2 n2+ b2
c •---= b ■ a2 + c2 - b2 = a2 + b2 - c2 bz = cz b = c.
2 ac 2 ab
Therefore, the triangle is isosceles with AB = AC, so the angles opposite these sides are equal: ZB = zC.
Solution 2. The equation AB cos B = AC cos C is equivalent to But by the Law of Sines, we have
= iiFL' so cole “ inC' which. means sin B cos C = cos B sin C. Then by the angle difference formula for sine, we
have
sin(B - C) = sin B cos C - cos B sin C = 0,
so B - C is an integer multiple of n. Since B and C are angles of a triangle, B - C must be equal to 0, so B = C.
72
Challenge Problems
4.46 From the Law of Sines, we have sin^BC = ^P, so sin/ABC = ^ sin LBCA = ^ • ± = 1. Therefore,
there is only one possible value of /.ABC, which is /ABC = 90°. This means that we must have /ABC = /DEB, so
A ABC = A DEE by AAS Congruence.
4.47 If /BAD — /CAD, then both angles are equal to A/2. Then by the Law of Sines on triangle A
ABD, we have ^ / and by the Law of Sines on triangle ACD, we have ^
But /ADB + /ADC = 180°, so sin /ADB = sin /ADC. Hence, we have 4M —
Challenge Problems
4.48 We have BD = tana and CD = tanp, so BC = BD + CD = tana + tanp. Also, cos a = AD/AB = l/AB so
AB = 1/ cos a. Similarly, we have AC = 1/ cos p.
The area of triangle ABC is given by \ AD • BC = |(tan a + tan j8). But the area of triangle ABC is also given by
1 111
-AB ■ AC sin /BAC = -- sin(a + p).
2 2 cos a cos p
l
Therefore, ^(tana + tanp) = \2 cos a cos /3
sin(a + p), which means
. , /sin a
sm(a + p) - cosacospftana + tanp) = cosacosp ^—— H-
sin B\ . „ . ^
= smacosp + smpcosa. -J
4.49 By the formula we found in the text for the cosine of half an angle of a triangle, we have
„ A B C s(s - a) s(s - b) s(s - a) 3(s - a)(s - b)(s - c) 4 ^s2 ■ s(s - a)(s - b)(s - c)
4 cos — cos — cos — = 4 =4
2 2 2 fee ac be a2b2c2 abc
The expression in the numerator looks like an expression in Heron's formula, which tells us that [ABC] =
Vs(s - a)(s - b)(s - c). We also showed in the text that abc = 4R[ABC]. Making these substitutions above gives
ABC 4 Js2 ■ s(s - a)(s - b){s - c) 4s[ABC] s

4 COS — COS — COS — = --- = = —.
2 2 2 abc 4B[ABC] R
4.50 Let Y and Z be the feet of the altitudes from B and C, as shown. From right A
triangle BHX, we have tan /HBX = ^, so BiX = BX tan /BiBX. We now seek information
about /BiBX and BX. From right triangle BYC, we have /HBX = /YBC = 90° - ZC, so
tan /HBX = tan(90° - ZC) = cotC. From right triangle AXB, we have cos B = ^|, so
BX = AB cos B. Finally, we have
cos C AB
HX = BX tan /HBX = BX cot C = AB cos B • —— = cos B cos C = 2R cos B cos C,
sm C sin C
where the Extended Law of Sines gives us the final step.
73
4.51 Let x = AC and 0 = iDCF. Then from right triangle ACF, we have cos 0 = A
CF/AC = l lx.
Since BD - CD, we have zDBC = iDCB = 0. Then zBDC = 180° - iDBC -

ZDCB = 180° - 20, so ZADB = 180° - ZBDC = 20. Then from right triangle BAD,
we have cos 20 = AD/BD = AD = AC - CD = x - 1.
By the double angle formula for cosine, we have cos 20-2 cos2 0-1. There¬
fore,
2
x - 1 = —z - 1 => X
2
= -r- =>
,
x= 2 => x= Hi
X1 X1
4.52 Extend AD and BC to intersect at E. Since ZEAB - lEBA = 60°, triangle ABE is
equilateral. Lets be the side length of equilateral triangle ABE. Then CE = BE-BC = s-8
and DE = AE - AD - s - 10.
By the Law of Cosines on triangle CDE,
(s - 8)2 + (s - 10)2 - 2(s - 8)(s - 10) cos 60° = 122

=> s2 - 16s + 64 + s2 - 20s + 100 - s2 + 18s - 80 = 144
^ s2 - 18s - 60 = 0.
By the quadratic formula.
18 ± Vl82 + 4 - 60 1 8±V564 18±2a/141

s = = 9 ± Vl41.
Since 9 - Vl41 < 0, we have s = 9 + Vl41
4.53 By the Pythagorean Theorem, the slant height of the cone is \/6002 + (200 V7)2 = \J(2002)(32 + (V7)2) =
200 V9 + 7 = 800. The circumference of the base is 2n ■ 600 = 120071, so when the curved surface of the cone is
unrolled, it becomes a sector with radius 800 and degree measure ^208°0q • 360° = 270°.
Let O be the center of the sector (which corresponds to the vertex of the cone), and
let A and B be the points where the ant starts and ends. Then the shortest path from
A to B on the cone corresponds to a straight line in the sector. Since A and B are on
opposite sides of the cone, we have /.AOB = 270°/2 = 135°. Therefore, by the Law of
Cosines,
x/2
AB2 = OA2 + OB2 - 20A ■ OB cos 135° = 1252 + (375 V2)2 + 2 • 125 • 375 V2 ~
= 1252(12 + (3 V2)2 + 2 • 1 • 3) = 56(25) = 58,
so AB = V51 = 625
4.54 By the angle difference formula for cosine, we have
cos(A -B) - cos A cos B + sin A sin B,
so cos A cos B = cos(A - B) - sin A sin B. Substituting into the given equation, we get
cos(A - B) - sin A sin B + sin A sin B sin C = 1,
so cos(A - B) = 1 + sin A sin B - sin A sin B sin C = 1 + sinAsinB(l - sinC).
74
Challenge Problems
Since A, B, and C are the angles of a triangle, we have 0 < sin A < 1, 0 < sin B < 1, and 0 < sin C < 1, so
1 + sinAsinB(l - sinC) > 1.
Furthermore, cos(A — B) < 1. Therefore, the equation above holds if and only if both sides are equal to 1, which
means sin A sin B(1 — sin C) = 0. Since sin A and sin B are positive, we have sin C = 1, which means C = 90°
4.55 Since 26 is an angle of a triangle, we have 0 < 26 < n, so 0 < 6 < f, which tells P
us that both sin 6 and cos 6 are positive. Let O be the center of square ABCD, and let M
be the midpoint of AB, so AM = AB/2 = 1/2.
Then by symmetry, ZPMB and ZPOM are right angles. Also, PM bisects LAPB, so
lAPM = 6. Then AM = PM tan 6, so
AM 1
PM =
tan 6 2 tan 6'
Now, OM = 1/2, so by the Pythagorean Theorem on triangle POM, we have
1 cos2 0 1 cos2 0 - sin2 0

OP2 = PM2 - OM2 =-V-
4 tan2 6 4 4 sin2 6 4 4 sin2 0
By the double angle formula for cosine, we have cos 20 = cos2 0 - sin2 0. Hence, we have OP2 = cos2 6
_ 4 sin 0
cos 20 __ Qp _ j cos 20 _ Vcos20
4 sin2 0' V 4 sin2 0 2 sin 0 ‘
Finally, by the formula for the volume of a pyramid, the volume of pyramid PABCD is
Vcos20 Vcos2 0
llABCD] ■ OP = i
2sin0 6 sin 0
4.56 Let O be the center of the circle. The measure of /BOC equals the measure
of BC in radians. Since the length of BC is r, and the circumference of the circle
is 2nr, the arc is ^ of the circle, which means its measure in radians is simply 1.
Therefore, we have ZBOC = 1 in radians. (Indeed, we can see this from the very
definition of a radian.) Then iBAC = BC/2 = /.BOC/2 = Since AB = AC, we
have ZABC = ZACB = (7z-|)/2=f-|. Then by the Law of Sines,
AB _ sin LACB _ sin(f - \) _ cos \

BC sin ZBAC sin | sin |
By the double angle formula for sine, we have sin \ = 2 sin | cos \, so
AB _ cos \ _ 1
BC 2 sin \ cos \ 2 sin |
4.57 Let the sides of the triangle be n - 1, n, and n + 1. Let the smallest angle be 0, so the largest angle is 20.
Therefore, the angle opposite n - 1 is 0, and the angle opposite n + 1 is 20, so by the Law of Sines,
n-l_n + l_ n +1
sin 0 sin 20 2 sin 0 cos 0
75
Solving for cos 0 gives cos 9 = 2(n-ij
By the Law of Cosines, we have
n2 + (n + l)2 - (ft - l)2 ft2 + ft2 + 2ft + 1 - ft2 + 2ft - 1 ft2 + 4ft _ ft + 4
cos 9 =
2 ft(ft +1) 2 ft(ft + 1) 2 ft(ft + 1) 2(ft + 1)
Setting this equal to our earlier expression for cos 9 gives
ft + 1 ft + 4 , i\, .c
2(^j = 2^TT) => (» + !) =(«-l)(n + 4) ft2 + 2ft + 1 = ft2 + 3ft - 4 ft = 5,
so the cosine of the smallest angle is cos 0 = ^ | = ||j-
4.58 Let x = CD. By the Angle Bisector Theorem, we have
CD _BD AB-CD _ 6x
= — => BD = = 2x.
AC ~ AB AC “ 3~
Let d = AD. By the Law of Cosines on triangle ACD, we have
32 + d2 - 2 ■ 3 • d cos 60° = x2 9 + d2 - 6d - \ = x2 d2 - 3d + 9 - x2.

2
and by the Law of Cosines on triangle ABD,
1
62 + dz - 2- 6 ■ dcos60° = (2x)2 => 36 + d2 - 12d ■ - = 4x2 d2 - 6d + 36 - 4x2.
Substituting the first equation into the second equation, we get
d2 - 6d + 36 = 4{d2 -3d+ 9) => d2-6d + 36 = 4d2 - 12d + 36 => 3d2 - 6d = 0 3d(d - 2) = 0.
Since d must be positive, we have d = |_2
As an extra challenge, solve the problem without using the Law of Cosines or the Angle Bisector Theorem.
4.59 We wish to show that AB > 2BC. We can write this as ^ > 2. So, P
our goal is to find an expression for ^ in terms of 9, and then show that
this expression is less than 2.
We see that iPBA = 180° - iPBC = 180° - 29, so lAPB = 180° -

LPAB - IPBA = 180° -9- (180° - 29) = 9. Similarly, LPCB - 180° - 39,
so LBPC = 180° - LPBC - IPCB = 180° -29- (180° - 39) = 9. Since
/.APB - /PAB = 9, we have AB = PB.
By the Law of Sines on triangle PBC, we have
PB _ sin /PCB _ sin(180° -30) _ sin 30

BC sin /BPC sin 0 sin 0 '
AB _ PB _ sin 30 /
BC BC sin0
By the triple angle formula for sine, we have sin 30 = 3 sin 0-4 sin3 0, so
AB 3 sin 0-4 sin3 0

= 3 - 4 sin2 0.
BC sin 0
76
Challenge Problems
Since 30 is an angle of a right triangle (other than the right angle), we have 30 < 90°, so 0 < 30°. And since the
sine function is increasing and positive on the interval [0, |], we have 0 < sin 0 < sin 30° = \. Therefore,
^ =3-4 sin2 0 > 3 - 4 = 2,

DC
so AB > 2BC.
4.60 Let m = BD and n = CD, as shown in the diagram at right (the diagram is not to A
scale). Then m + n = a, and by the Angle Bisector Theorem, we have m/c = n/b, so bm = cn.
Substituting n = a-m, we get bm = c(a - m). Solving for m, we find
ac
m =
b + c'
so
bm ab
n=
b+c
Let t = AD. By Stewart's Theorem, we have a(t2 + mn) = mb2 + nc2. Substituting for m and n, the left side
becomes
a2bc
a(t + mn) = a [t +
(b + cf
and the right side becomes
,9 9 ab2c + abc2 abcCb + c) ,
mb + nc - —;-= —r--= abc,
b+c b+c
so a (f2 + ) = abc, which means t 2 + (LkT = kc■ Isolating t2, we find
a2bc bc(b + c)2 - a2bc _ bc((b + c)2 - a2)

r - be -
(b + c)2 (b + c)2 (b + c)2
\Jbc((b + c)2 - a2)

so t =
b+c
4.61 Solution 2. Let 0 = ZBAM, so ZBDM - 30. From right triangle BAM, we have
tan 0 = fg = yy, and from right triangle BDM, we have tan 30 = = BM.
Therefore, we have = 11. We also have
tan 20 + tan 0 + tan 0

1-tan2
n * r»z f61
tan 30 = tan(20 + 0) = 2 tan2 6

1 - tan 20 tan 0
1-tan2 9
Multiplying the numerator and denominator of this last expression by 1 - tan2 0 gives
2 tan 0 + tan 0(1 - tan2 0) _ 3 tan 0 - tan3_0 C

tan36= 1-3 tan2 0 " 1-3tan2 0
Therefore, our earlier equation =11 gives us
tan 30 3 - tan2 0
11 =
tan 0 1-3 tan2 0
Solving for tan 0, and noting that tan 0 > 0 since 0 is acute, we have tan 0 = 1. This gives us BM = f, and the
Pythagorean Theorem applied to AAMB gives AB = so the perimeter of triangle ABC is AB + AC + BC =
2AB + 2 BM = 11 + 11 V5
77
Solution 2. Let 0 = zBAM. Then zBAC = 2zBAM = 29, LBDC = 3LBAC = 60, and ZBDM = ZBDC/2 = 30.
Then ZBDA = 180° - ZBDA4 = 180° - 30, so ZABD = 180° - ZBAD - ZBDA = 180° - 0 - (180° - 30) = 20.
Let x = BD. By the Law of Sines on triangle ABD,
BD _ AD x _ 10 _ 10 _ 5
sin ZBAD sin ZABD sin 0 sin 20 2 sin 0 cos 0 cos 0 sin 0'
so cos 0 = |.
AM _ 11 _ llx
Then from right triangle ABM, we have AM = AB cos 0, so AB = cos0 5/x 5
By the Pythagorean Theorem on right triangle ABM, we have
121 (x2 - 25)

BM2 = AB2 - AM2 = - ll2 =
25
By the Pythagorean Theorem on right triangle DBM, we have BM2 = BD2 - DM2 = x2 - 1. Hence,
121 (x2 - 25) =xz_1 ^ 121(x2 - 25) - 25 (x2 - 1) => x=

5V5
2 '
Then AB = ^ = L\fi, and BM = Vx2 -1 = = = y, so the perimeter of triangle ABC is
AB+AC + BC = 2AB + 2 BM = 11 + 11 V5
4.62 Solution 1: Geometry. We must create a diagram with an 18° angle,

but where in geometry might we encounter such angles? We note that
18° = ^ radians. The factor of 5 in the denominator suggests investigating
a regular pentagon, which we have drawn at right. We'd like to form a right
triangle that we can use to find sin 18°. We can build such a right triangle
by drawing altitude DG from D to AB, which bisects ZEDC. We have B
ZEDG = 54°, since each interior angle of a regular pentagon has measure
108°. Then, drawing AD forms isosceles triangle AED in which we have
ZADE = ZEAD - (180° - zE)/2 = 36°, so ZADG = 18°. We now have a right
triangle, AADG, with one angle equal to 18°. So, if we let AD = 1, then we
seek the length of AG, since sin 18° =
We let AG = x. We build another right triangle with one angle equal

D C
to 18° by drawing altitude BH from B to AD. We have ABHA ~ ADGA by
AA, since these two right triangles share an acute angle. We have AB = 2AG - 2x. We can introduce yet another
18° angle in a right triangle by drawing BE. Letting F be the intersection of BE and AD, we see that BFDC is a
parallelogram. Furthermore, since BC = CD, this parallelogram is a rhombus, and we have BF = ED = CD =
AB = 2x. Since BF = AB, altitude AH of isosceles triangle ABF bisects AF, so AH = 4f = ^--yP = ^y^ = — x. \
Returning to similar triangles ABH and ADG, we have jj=j = so j = —. Cross-multiplying gives
2x2 - \ - x. Multiplying by 2 and rearranging gives 4x2 + 2x - 1 = 0. The quadratic formula then gives
-1+V5
x= 2± Y )+16 =
2 4 l\^■ Since x must be positive, we have sin 18° = x =
Solution 2: Algebra Let x = cos 36° and y = cos 72° = sin 18°. Then by the double angle formula for cosine, we
have cos 72° = 2 cos2 36° - 1, so
y = 2x2 - 1.
Again by the double angle formula for cosine, we have cos 36° = 1 — 2 sin2 18° = 1-2 cos2 72°, so
x = 1 - 2y2.
78
Challenge Problems
Adding these equations, we get x + y - 2(x2 - y2) - 2(x + y)(x - y). Both x and y are positive, so x + y is positive.
Hence, we may divide both sides of x + y = 2(x + y)(x - y) by x + y to get 2(x - y) = 1. Solving for x, we find
x ~ V + t Substituting into the equation x = 1 - 2y2, we get y + \ — 1 — 2y2, which simplifies as 4y2 + 2y - 1 = 0.
By the quadratic formula, we have y = Since < 0, we conclude sin 18° = -1+V5
4
4.63 First, to evaluate the given expression, we express all the trigonometric functions in terms of sine and cosine:
cosy
c°t}/ _ siny _ sin a sin jS cos y
cot a + cot f$ cos a + cos fi sin y (sin a cos + cos a sin jS) ’
sin a sin j8
By angle sum formula for sine, we have sinucos/3 + cos a sin j8 = sin(« + jS) Furthermore, a + fi — tl — y, so
sin(a + jS) = sin(7i - y) = sin y. Therefore,
sin a sin j8 cos y sin a sin fi cos y

sin y (sin a cos jS + cos a sin /?) sin2 y
By the Law of Sines,7 we have -A— b „„ sin a _ a j sin ft _ b

' sma sin ft siny' siny “ c anCl sin y c . Hence,
sin a sin jS cos y sin a sin B ab

—— • cos y = — cos y.
sin2 y sin y sin y c2
Finally, by the Law of Cosines, ^ cos y = ^ c2 — g2+fo2 g2 994

2c2 2c2
4.64 We have tan y = |~, tan | = and tan ~ so
A B AC B C r r r r r
tan — tan - + tan — tan - + tan - tan - = + — --1-- _ r _ ^(s - c) + r2(s -b) + r2(s - a)
s- a s-c s—b s-c (s - a)(s - b)(s - c)
r2[3s -(a + b + c)] r2(3s - 2s) tAs
(s - a)(s - b)(s - c) "(s - a)(s - b)(s - c) (s - a)(s - b)(s - c)
r2s2 _ [ABC]2
1.
s(s - a)(s - b)(s - c) [ABC]2
4.65 We start with a cross-section of the Earth that includes Vancouver

and the Earth's axis. In the diagram, Vancouver is point V and O is the
center of the Earth. The Sun's rays come in from point S, putting half
the Earth in night and half the Earth in daytime. b(3 is the dividing line
between night and day, so it is perpendicular to the Sim's rays. CO is
the axis of the Earth, point E is on the equator of the Earth, and VY is
Vancouver's line of latitude (which means that V and Y are the same
distance from the equator, so VY is perpendicular to the Earth's axis).
The length of daytime in Vancouver is maximized when we maximize

the portion of VY that is in the "Day" section of the diagram. The axis of
the Earth always makes a roughly 66.5° angle with the plane in which it
orbits the Sun (we're simplifying here—the Earth "wobbles" a bit about this axis). However, as the Earth revolves
about the sun, the angle between the Sun's rays and the upper half of the Earth's axis varies.
79
To get some intuition for why, consider the diagram above. The Sun is at the center of the diagram, and two
positions of the Earth are shown. On the left is the Earth when Vancouver is in winter, and the days are short.
On the right is Vancouver in summertime, when the days are long. The Earth's axis always points in the same
direction (roughly at the "North Star") as it rotates around the Sun, so the axes in the two Earth positions shown
are parallel.
The smallest that ZCOS can be is when the upper half of the Earth's axis points towards the Sun, and the
largest it can be is when this upper half of the axis points away from the Sun. Since the axis makes an angle of
66.5° with the plane in which the Earth orbits, this means that ZCOS ranges from 66.5° to 180° - 66.5° - 113.5°. As
our diagram suggests, the smaller ZCOS is, the more of Vancouver's latitude is included in daytime. Therefore,
summer solstice, the longest day of the year in Vancouver, occurs when ZCOS = 66.5°.
Now we're ready to do some geometry. When the Earth rotates about its axis, Vancouver
will move along its line of latitude one full turn about the axis. We wish to know for what
portion of this trip Vancouver will be in sunlight. To measure this, we consider a second
cross-section of the Earth, this time along Vancouver's latitude. This cross-section is at
right. Points C and D are the same as in our initial diagram: C is where the axis of the
Earth intersects the plane of Vancouver's latitude, and D is where the line that divides
night and day in our first diagram intersects the line along Vancouver's latitude in that
diagram. (Picture looking at the Earth from directly above the North Pole—C is on the axis
and therefore at the center of your view, and D is the midpoint of the segment that divides night and day.) Points
A and B are the boundary points between night and day.
To determine how long Vancouver is in sunlight, we must determine what portion of the circumference of this
cross-section is in sunlight. To do this, we must find ZACB. We start by drawing altitude CD to AB, both because
this forms right triangles and because we can learn more about CD from our first cross-section. Right triangles
ACD and BCD are congruent, so our problem now is to determine ZACD, which we let be a. We can find a by
first finding two side lengths of A ACD, and then using trigonometry.
To find these side lengths, we return to our initial cross-section. CD

is the same in this cross-section as in the latitude cross-section. CA in
the latitude cross-section connects a point on Vancouver's latitude to the
nearest point on the Earth's axis. Therefore, it has the same length as CV
in the cross-section at right. Now, our problem is to determine CD and
CV. Our diagram has many right triangles, and we know a lot of angles,
so we reach for our trigonometric tools.
Because ZCOS = 66.5°, we have ZCOD = 90° - 66.5° = 23.5°. Since

the latitude of Vancouver is 49° N, we have ZVOE = 49°, which means
that ZCOV = 90° - ZVOE = 41°. Letting R be the radius of the Earth,
we have VO - R. We now focus on the right triangles that contain our
target lengths, CD and CV. Right triangle COV gives us sin ZCOV = so CV = VO sin ZCOV - R sin 41°. In
80
Challenge Problems
order to use A COD to build an expression for CD, we need an expression for CO or OD. Fortunately, we can use
ACOV to tackle CO. We have cos ZCOV - so CO = VOcos/COV = Kcos41°. Finally, in A COD, we have
tan /.COD = so CD = CO tan ZCOD = P cos 41° tan 23.5°.
We're now ready to return to the latitude cross-section. Since CA in this cross-section
is the same as CV in the axis cross-section, we have CA = R sin 41°. We also found that
CD = R cos 41° tan 23.5°, so we can write an expression for a:
CD R cos 41° tan 23.5°

cos a - 0.500.
CA R sin 41°
(Yes, now you can see why we chose Vancouver for this example!) Since a is acute, cos a « 0.5
tells us that a ~ 60°. Therefore, ZACB ~ 120°, which means that \ of the circumference of
the circle is in the "night" section. This leaves the other § of the circle in daytime, and gives Vancouver §(24) = 16
hours of sunlight on the summer solstice.
4.66 By the Pythagorean Theorem, we have AB - 25. Since M is

the midpoint of hypotenuse AB, M is the circumcenter of triangle
ACB, so CM = x = ¥ • Since aADB is isosceles with AD - DB and
point M is the midpoint of AB, we have ZBMD = 90°. Applying the
Pythagorean Theorem to ADMB gives
275 _ 5 Vll
DM = y/BD2 - BM2 =
T~ ~ 2
Let 0 = /CMD, so /CMB - /CMD + ZDMB = 0 + 90°. Then by the Law of Cosines on triangle BMC,
527
BM2 + CM2 -BC2 ^ + ^-576 - 527
cos(0 + 90°) =
IBM ■ CM 9
Z •
25
2 •
25
2
625'
Since cos(0 + 90°) = - sin 0, we find sin 0 = |||. Therefore, the area of triangle CMD is
1 „ 1 25 5 Vll 527 527 Vll

-CM-DMsm0=--y — 40
4.67 Let the circle be tangent to BC and DA at R and S, respectively, and let A P B
O and r be the center and radius of the circle, respectively. Let a = ZPOA,
= zPOB, y - ZQOC, and 6 = ZQOD. Because OP = OS = r, zAPO -
ZASO = 90°, and right triangles APO and ASO share hypotenuse AO, we have
AAPO = AASO by HL Congruence for right triangles. Therefore, we have
ZSOA = ZPOA - a. Similarly, /ROB = ZPOB = jS, ZROC = ZQOC = y, and
ZSOD = ZQOD = b. Summing the angles around O, we get 2a + 2fi + 2y + 2b =
2n, so a + p + y + b = n.
From right triangles POA and FOB, we have tana = y and tan^ = y.
tana + tanjS _ 19/r + 26/r _ 45r

tan(a + ft) ^ _ ân a tan ^ 1 — 19/r • 26/r r2 - 494
81
_ 23
Similarly, from right triangles QOC and QOD, we have tan y = y and tan [i = so
tan y + tan <5 _ 37/r + 23/r 60 r

tan(y + 6) =
l-tanytan5 l-37/r*23/r r2 - 851
Since (a + jS) + (y + 5) = 7i, we have tan(n + fi) + tan(y + 6) = 0. Therefore,
45r 60r
0 + 3(r2 - 851) + 4(r2 - 494) = 0 Ir2 = 4529,
r2 _ 494 + ——~
r2 _gs! =
r2 - 494 r2 - 851
so r2 = 647
4.68 By the double angle formula for cosine, we have
? „ cos 2B + 1 . 7 _ cos 2C + 1
cos B =--- and cos C =---.
so
o_ 7 _ cos 2B + 1 cos 2C + 1 cos 2B + cos 2C
COS B+ COS C = -r- + --- = --- + 1.
Then by the product-to-sum formula for cosine,
cos2B + cos2C „ .
---h 1 = cos(B + C) cos(B - C) + 1.
But B + C = 180° — A, which means cos(B + C) = cos(180° — A) — — cos A. Therefore,
cos(B + C) cos(B - C) + 1 = - cosAcos(B - C) + 1.
Hence,
cos2 A + cos2 B + cos2 C = cos2 A ~ cos A cos(B - C) + 1

= 1 — cosA[cos(B - C) - cos A]
= 1 - cos A[cos(B - C) - cos(180° - B - C)]
= 1 - cosA[cos(B - C) + cos(B + C)].
Finally, by the sum-to-product formula for cosine, cos(B — C) + cos(B + C) = 2 cos B cos C, so
1 - cos A[cos(B - C) + cos(B + C)] = 1 — 2 cos A cos B cos C.
Hence, cos2 A + cos2 B + cos2 C = 1 - 2 cos A cos B cos C, or cos2 A + cos2 B + cos2 C + 2 cos A cos B cos C = 1.
4.69
(a) Let a = AB, b = BC, c = CD, and d = DA. By the Law of Cosines on triangles BAD and B
BCD, we have BD2 = a2 + d2 - 2ad cos A and BD2 - b2 + c2 - 2be cos C, respectively. Since
quadrilateral ABCD is inscribed in a circle, we have C = 180°-A, so cos C = cos(180°-A) =
gives
a2 + d2 - 2fld cos A = bz + c2 + 2bc cos A a2 -b2 - c2 + d2 = 2(ad + be) cos A
a2 -b2 -c2 + d2
cos A =
2 (ad + be)
Then
i2 - b2 - c2 + d2 n 2 [2(ad + be)]2 - (a2 -b2 - c2 + d2)2

sin2 A = 1 - cos2 A = 1
2 (ad + be) [2 (ad + be)]2
[2(ad + be) + (a2 -b2 - c2 + d2)][2(ad + be) - (a2 -b2 -c2 + d2)]
[2 (ad + be)]2 (♦)
82
Challenge Problems
The first factor in the numerator is equal to
2 (ad + be) + (a2 -b2 - c2 + d2) = (a2 + 2 ad + d2) - (b2 - 2 be + c2) - (a + d)2 - (b - c)2 = (a + d + b- c)(a + d-b + c).
Since 2s = a + b + c + d, we have a + d + b - c - (a + b + c + d) - 2c = 2s - 2c = 2(s - c), and a + d- b + c =

(a + b + c + d) - 2b - 2s — 2b — 2(s - b), so
(a + d + b - c)(a + d - b + c) = 4(s - b)(s - c).
Similarly, the second factor in the numerator of (*) is equal to
2 (ad + be) - (a2 -b2 -c2 + d2) = (b2 + 2 be + c2) - (a2 - 2 ad + d2) = (b + c)2 - (a- d)2
= (b + c + a - d)(b + c- a + d)
-[(a + b + c + d) - 2 d][(a + b + c + d) - 2a] - (2s - 2d)(2s - 2a) - 4(s - d)(s - a),
so
2 A _ 16(s - a)(s - b)(s - c)(s - d) _ 4(s - a)(s - b)(s - c)(s - d)
sirr —-———;—:——r-—--—-—-——-
[2 (ad + be)]2 (ad + be)2

Since 0° < A < 180°, sin A is positive, so
_ 2 V(s - a)(s - b)(s - c)(s - d)

ad + be
Quadrilateral ABCD is composed of the two triangles BAD and BCD. The area of triangle BAD is
1 1
[BAD] = -AB ■ AD sin LB AD = -ad sin A.
The area of triangle BCD is
[BCD] = ^BC • DC sin LBCD - ^bc sin C.
But sinC = sin(180° - A) = sin A, so [BCD] = ^bc sin A. Therefore,
[ABCD] - [BAD] + [BCD] = -ad sin A + -be sin A = -(ad + be) sin A = y/(s- a)(s - b)(s - c)(s - d).
(b) By the Law of Cosines on triangles BAD and BCD, we have BD2 - a2 + d2 - 2ad cos A and BD2 = b2 + c2 -
2be cos C, respectively, so
a2 + d2 - 2ad cos A = b2 + c2 - 2be cos C.
Then a2 -b2 - c2 + d2 - 2ad cos A - 2be cos C. Squaring both sides, we get
(a2 -b2 - c2 + d2)2 = 4a2d2 cos2 A + 4b2c2 cos2 C - 8abcd cos A cos C. (*)
Quadrilateral ABCD is composed of the two triangles BAD and BCD, so
1 1
[ABCD] = [BAD] + [BCD] = -ad sin A + -be sin C,
or 4[ABCD] = 2ad sin A + 2be sin C. Squaring both sides, we get
16[ABCD]2 = 4a2d2 sin2 A + 4b2c2 sin2 C + 8abed sin A sin C. (**)
Adding equations (*) and (**), we get
16[ABCD]2 + (a2 -b2 -c2 + d2)2 = 4a2d2(cos2 A + sin2 A) + 4b2c2(cos2 C + sin2 C) - 8abcd(cos A cos C - sin A sin C)
= 4a2d2 + 4b2c2 - 8abcd(cos A cos C - sin A sin C).
83
By the angle sum formula for cosine, we have cos A cos C - sin A sin C = cos(A + C), and by the double angle
formula for cosine, we have cos (A + C) = 2 cos2 - 1, so
2 j2
4a2d2 + 4b2c2 - 8abcd(cos A cos C - sin A sin C) = 4azdz +4 ,
au2„2
bzcz - 8abcd 2 cos2
A+C
-1
A+C
= 4a2d2 + 8abed + Ab2 c2 - 16abed cos2
A+C
= [2(ad + be)]2 - 16abed cos2
Hence,
A+C
16[ABCD]2 + (a2 -b2 - c2 + d2)2 = [2(ad + be)]2 - \6abcd cos2
which implies that
A+C
16[ABCD]2 = [2(ad + bc)]z - (a12 -bz
u2 ~2
- cl + dz)z - 16abcd cos
We have already calculated in part (a) that
[2(ad + be)]2 - (a2 -b2 - c2 + d2)2 = 16(s - a)(s - b)(s - c)(s - d),
so
A+C
16[ABCD]2 = 16(s - a)(s - b)(s - c)(s - d) - 16abed cos2
Then [ABCD]2 = (s - a)(s - b)(s - c)(s - d) - abed cos2 (^^), so
A+C
[ABCD] = \ (s - a)(s - b)(s - c)(s - d) - abed cos2
Note: This is called Bretschneider's formula, after someone who must have really liked algebra.
4.70 Let 6 - IDBA. Then /CAB = iDBC = 26. Hence, ZAOB = A B

180° - ZOAB - lOBA = 180° -26- 9 = 180° - 30, so LBOC =
180° - LAOB = 30.
Then by the Law of Sines on triangle BOC, we have ^ =

1^11, and by the Law of Sines on triangle ABC, we have =
sin 38
sin 28 ‘
But quadrilateral ABCD is a parallelogram, so O is the midpoint of AC, which means AC = 20C. Therefore,
sin 30 _ AC _ 20C _ 2 sin 20

sin 20 BC BC sin 30
so 2 sin2 20 = sin2 30.
By the double angle formula for sine, we have sin 20 = 2 sin 0 cos 0, and by the triple angle formula for sine,
we have sin 30 = 3 sin 0 - 4 sin3 0 = sin 0(3 - 4 sin2 0). Therefore,
8 sin2 0 cos2 0 = sin2 0(3 - 4 sin2 0)2.
Since sin 0 is not equal to 0, we may divide both sides by sin2 0, to get
8 cos2 0 = (3 - 4 sin2 0)2.
84
Challenge Problems
By the double angle formula for cosine, we have cos2 0 = (1 + cos 20)/2 and sin2 0 = (1 - cos 26)12, so
4 + 4 cos 26 - (1 + 2 cos 20)2.
Expanding the right-hand side and simplifying, the equation above becomes 4cos2 26 = 3, so cos2 26 =
Since 26 and 36 are angles of triangle BOC, we have 56 < 180°, so 6 < 36°. Then 20 < 72°, so 20 is acute,
which means cos 20 is positive, so cos 20 = Hence, 20 = 30°, so 0 = 15°.
Finally,
lACB 180° - 50 _ 105° _ |~7~
' " LAOB ~ 180° -30 “ 135° “ 9
4.71 Without loss of generality, assume that ZB > ZC. (The case ZB < ZC is tackled A
similarly.) The differences of angles in the formula make us look for an angle with
measure B - C. We construct one as follows: Let D be the foot of the altitude from A
to side BC, and let E be the reflection of B through D. This creates another angle with
measure B, since AE = AB implies that ZAEB = ZB. Now, since ZAEB is an exterior angle
of AAEC, we have ZAEB = ZC + lEAC. Therefore, lEAC = lAEB - 1C = IB - ZC.
Turning to side lengths, from right triangle ABD, we have BD = AB cos lABD = c cos B,
and from right triangle ACD, we have CD = AC cos lACD - bcosC. Then DE = BD =
c cos B, so CE = CD - DE = b cos C - c cos B. Then by the Law of Cosines,
+ b2 a2 + c2 — fo2 a2 + b2 - c2 a2 + c2 - b2 2b2 - 2c2 b2 - c2

CE - b cos C - c cos B - b ■ -c•
2ab 2 ac 2fl 2a 2fl a
Since AE = AB = c and AC = b, applying the Law of Cosines to triangle ACE gives
AC2 + AE2 - CE2 fr2 + c2 - (^)2 _ a2fc2 + a2c2 - (b2 - c2)2

cos(B - C) = cos zEAC =
2AC-AE 2 be 2 a2bc
a2b2 + a2c2 — fc4 + 2b2c2 - c4
2a2bc
Hence,
a3(a2b2 + a2c2 + 2b2c2 -b4- c4) a2(a2b2 + a2c2 + 2b2c2 - b4 — c4) a4b2 + a4c2 + 2a2b2c2 — a2b4 — a2c4
a3 cos(B-C)
2a2bc 2abc 2abc
Similarly,
a2!?4 + l?4c2 + 2a2b2c2 - a4b2 - b2c4

b3 cos(C - A) =
2abc
a2c4 + b2c4 + 2a2b2c2 - a4c2 - b4c2
c3 cos(A - B)
2 abc
Adding all three equations, we get
6a2b2c2
a3 cos(B - C) + b3 cos(C - A) + c3 cos(A - B) = 3afc>c.
2 abc
85
CHAPTER 5. PARAMETERIZATION AND TRIGONOMETRIC COORDINATE SYSTEMS
CHAPTER 5 i
Parameterization and Trigonometric Coordinate Systems
5.1.1 We eliminate t from the two equations. Multiplying the equation for x by 8 and the equation for y by 3
gives 8x = 241 + 56,3y = 241 — 15. Subtracting the second equation from the first gives 8x - 3y = 71. Solving for y
gives y - |x - y. The slope of the graph of this equation is
5.1.2 We have x2 + y2 = 9 cos2 2t + 9 sin2 21 = 9(cos2 21 + sin2 2t) = 9. Therefore, all points on the graph are on the
circle centered at the origin with radius 3. As t varies from 0 to n, every point on this circle is produced, so the
graph of the parametric equations is the circle centered at the origin with radius 3
5.1.3 From the given equations, we have x + 3 = 5 cos 0 and y - 2 = 5 sin 0. Squaring both and adding them gives
(x + 3)2 + (y - 2)2 = (5 cos 0)2 + (5 sin 8)2 = 25 cos2 0 + 25 sin2 0 = 25.
Thus, the graph is a circle with center (-3,2) and radius 5 . We can see that the graph of the parametric equations
is the entire circle as follows. First, we note that the graph of x = 5 cos 0, y = 5 sin 0 is a dilation of the unit circle
about the origin, so the graph of x = 5 cos 0, y = 5 sin 0 is a circle centered at the origin with radius 5. Each point
on the graph of x = -3 + 5 cos 0, y = 2 + 5 sin 0 is 3 to the left and 2 above the corresponding point on the graph
x = 5 cos 0, y = 5 sin 0. So, the graph of x = -3 + 5 cos 0, y = 2 + 5 sin 0 is the result of translating the graph of
x = 5 cos 0, y = 5 sin 0 by 3 to the left and 2 up. Therefore, the graph is a full circle.
5.1.4 Solving x = 4cos2f for cos 21 gives cos 21 = |. Substituting this into
y = 1 + cos2 2t gives y = 1 + fy So, all points on the graph of the parametric
9
equations are on the graph of y = 1 + fg. However, the parametric equations

9
do not produce the entire parabola described by y = 1 + ^. Since cos 21 varies

from -1 to 1, the value of x = 4 cos 21 varies from -4 to 4. Therefore, graphing
the parametric equations only pro*
produces the portion of the graph of y = 1 + ^
with -4 < x < 4, as shown at right.
5.1.5 Dividing both sides by 144 gives fg + = 1. The graph of this equation is an ellipse. To parameterize it,
we notice that the two terms on the left are perfect squares whose sum is 1. Therefore, we can take ^ = cos2 0 and
|/2 r\
fj. = sin 0. One solution for x and y is x = 6 cos 0, y = 4 sin 0 . Graphing these parametric equations as 0 ranges
from 0 to 2tz traces out the full ellipse.
These are not the only parametric equations whose graph is this ellipse. We could also have taken x =
-6 cos 0, y = 4 sin 0, or x = -6 cos 20, y = -4 sin 20, etc. We can find infinitely many other possibilities. We can let
86
Section 5.1
x be 6 or -6 times either sine or cosine of some angle, and let y be 4 or -4 times the cosine or sine (we choose the
trig function we didn't use for x) of the same angle we used to express x.
5.1.6 We can parameterize the unit circle with x = cos t, y - sin t, so we can parameterize a circle with radius 3
centered at the origin with x = 3 cos t, y = 3 sin t. The circle with radius 3 centered at (4,5) is a translation right
4 units and up 5 units of the circle with radius 3 centered at the origin. Therefore, if (x, y) is a point on the circle
centered at (4,5), then there is some point (x', y') on the circle centered at the origin such that x = 4 + x', y = 5 + y'.
We can parameterize the circle centered at the origin with x' = 3 cos t, y' = 3 sin t, so the circle centered at (4,5)
with radius 3 can be parameterized with x = 4 + 3 cos t, y = 5 + 3 sin t.
While the graph of the parameterization x = 4 + 3 cos t, y = 5 + 3 sin t does produce the same circle that the
particle traces, it does not accurately describe the location of the particle. At t = 0, the equations x = 4 + 3 cos t,
y = 5 + 3 sin t give (x, y) = (7,5), but the particle is at (1,5), which is diametrically opposite (7,5). We can account
for this by introducing a phase shift of n, using the parameterization x = 4 + 3cos(f + n), y = 5 + 3sin(f + n).
Applying the trig identities cos(f + 7i) = — cos t and sin(t + tt) = — sin t makes this parameterization x = 4 - 3 cos t,
y = 5 - 3 sin t.
We're not quite finished yet. We check if the the particle is moving in the right direction by noting that if t = |,
then x = 4 - 3 cos t, y = 5 - 3 sin t gives (x, y) = (4,2), which is indeed counterclockwise from (1,5). However, we
have to take into account the speed of the particle. Since it travels around the whole circle in 871 seconds, we need
the period of each of the trig functions to be 8n rather than 2n. We accomplish this by replacing t with t/4, and
t t
we have the parameterization x — 4 — 3 cos y = 5 - 3 sin -
4 4
5.1.7 We can't easily eliminate the parameter, but we can at least eliminate the denominators of the fractions.
Since x clearly cannot be 0, we can divide our equation for y by our equation for x, which gives \ = t. Aha! Now,
we can substitute for the parameter! Substituting into the equation for x gives
4 _ 4 _ 4x2
= 77Z = T+tf-'
Since x is nonzero, we can divide both sides by x, and multiply both sides by x2 + y2, to get x2 + y2 = 4x. The graph
of this equation is a circle. We rearrange the equation and complete the square to find that (x - 2)2 + y2 = 4. The
graph of this equation is a circle with center (2,0) and radius 2.
We can't yet conclude that the graph of the parametric equations is this circle. We have
only shown that every point on the graph of the parametric equations is on this circle.
Looking at the equation x = we see that the largest x can be is 4, which occurs when
t = 0. For any other f, we have 1 + t2 > 1, so x < 4. We also see that x must be positive, and
can take on any value in the interval (0,4]. Specifically, we see that we must omit the origin
from the graph, so the graph of the parametric equations is not the entire circle centered at
(2,0) with radius 2.
We're still not quite finished—for each value of x such that 0 < x < 4, there are two points on the circle. Do our
parametric equations produce both? Yes! The one above the x-axis is produced when t > 0 and the one below the
x-axis is produced when t < 0. Therefore, the graph of the parametric equations is the circle centered at (2,0) with
radius 2, except the origin, as shown at right above.
87
5.2.1
(a) We have x = -5 and y = 0, so r = yjx2 + y2 = yj(-5)2 + 02 = V25 = 5. We seek an angle 0 such that
5 cos 0 = -5 and 5 sin 0 = 0, which means cos 0 = -1 and sin 0 = 0. We can take 0 = 180°. Hence, (5,180°)
are polar coordinates of the point.
(b) We have x = 6 V2 and y = 6, so r = yjx2 + y2 = \j(6 V2)2 + 62 = Vl08 = 6 V3. We seek an angle 0 such that
6 V3 cos 0 = 6 V2 and 6 V3 sin 0 = 6, which means cos 0 = and sin 0 = -^. Since the point (6 V2,6) lies
in the first quadrant, we can take 0 = arccos y 5 ~ 35.3°. Hence, (6 V3,35.3°) are polar coordinates of the
point.
(c) We have x = -4 and y = -4 V3, so r = yjx2 + y2 = \/(-4)2 + (-4 V3)2 = V64 = 8. We seek an angle 0 such
_ V3
that 8cos 0 = -4 and 8 sin 0 = -4 V3, which means cos 0 = and sin 0 . We can take 0 = 240°.
Hence, (8,240°) are polar coordinates of the point.
5.2.2
(a) We have r = 7 and 0 = 210°, so x = 7cos 210° = 7- (-4^) = and y = 7sin210° = 7- (—|) = — Hence,
the rectangular coordinates are (-¥--!)

(b) We have r = 9 and 0 = -180°, so x = 9cos(-180°) = 9 • (-1) = -9, and y = 9sin(-180°) = 9-0 = 0. Hence,
the rectangular coordinates are (-9,0)
(c) We have r = -8 and 0 = so x = -8 cos = -8 • = -4 V2, and y = -8 sin ^ = -8 • = —4 V2. Hence,
the rectangular coordinates are (-4 V2, -4 V2)
5.2.3 Multiplying both sides by 2 cos 0-5 sin 0, we get 2r cos 0 - 5r sin 0 = 4. We don't have worry about
introducing extraneous solutions here, since clearly this resulting equation has no solutions for which we have
2r cos 0 - 5r sin 0 = 0. Since x = r cos 0 and y = r sin 0, this equation becomes 2x - 5y = 4. Thus, the graph is a
line . (This line has x-intercept (2,0) and y-intercept (0, -§)•)
5.2.4 We can express r = 6 sin 0 in the form given for a circle in the text,
r2 - 2rro cos(0 - a) + r\ = f2,
with a little clever manipulation. We multiply both sides of r = 6 sin 0 by r and move everything to the left side to
get r2 - 6r sin 0 = 0. Then, we note that sin 0 = cos(90° - 0) = cos(-(90° - 0)) = cos(0 - 90°), so we can write the
equation r2 - 6r sin 0 = 0 as r2 - 6r cos(0 - 90°) = 0. Almost there! Comparing this to our form given for a circle,
we see that we can let ro = 3 and a = 90°. Finally, we must also have t = 3 to make the constants cancel in the
form r2 - 2rro cos(0 - a) + = t2. So, we write our equation as:
r2 - 2(3)rcos(0 - 90°) + 32 = 32.
We thereby see that the graph of r - 6sin0 is a circle with radius 3 and center (3,90°) (which is indeed (0,3) in
rectangular coordinates).
5.2.5
(a) For segments AO and BO to be perpendicular, the angles 0 and 137° must differ by an odd multiple of 90°.
Therefore, the possible values of 0 are 137° + 90° 227° and 137° - 90° = 47°
88
Section 5.2
(b) If AB _L AO, then ZOAB = 90°. Then cos lAOB = gf = £ = \, so lAOB = 60°. Therefore, the possible
values of 6 are 137° - 60° = 77° and 137° + 60° = 197°
5.2.6 When we graph r = 2 + 39, r increases as 6 increases, so the graph gets farther and
farther from the origin as 6 increases. Because increasing 6 when r > 0 means rotating
around the origin, we obtain a spiral as the graph. The graph of an equation of the form
r = a + bd is called an Archimedean spiral.
5.2.7 First, we determine the polar coordinates of the point (3,3). We have x - 3 and y = 3,
so r = \Jx2 + y1 - V32 + 32 = Vl8 = 3 V2. We seek an angle 6 such that 3 V2 cos 0 = 3 and
3 V2 sin 0 = 3, which means cos 0 = -y and sin 0 = We can take 0 = 45°. Hence, the
polar coordinates are (3 V2,45°).
To find the other vertices, we note that if we draw a circle centered at the origin such that the circle passes
through the six vertices of the hexagon, then the vertices are equally spaced about the circle. So, in polar
coordinates, the radius is the same for each, and we find the angle of successive vertex by adding 360°/6 = 60°
to the angle of the previous vertex. Then the angles corresponding to the other five vertices of the hexagon are
45° + 60° = 105°, 105° + 60° = 165°, 165° + 60° = 225°, 225° + 60° = 285°, and 285° + 60° = 345°. The two second
quadrant angles are 105° and 165°.
Now we must compute the coordinates. We have:
cos 15° = cos(45° - 30°) = cos 45° cos 30° + sin 45° sin 30° = ^ ^ ~~, and
a/6- V2
sin 15° = sin(45° - 30°) = sin 45° cos 30° - sin 30° cos 45° = ■—.
Then
-V6 + V2
cos 105° = cos(15° + 90°) = - sin 15° =
4
V6 +V2
sin 105° = sin(15° + 90°) = cos 15° =
cos 165° = cos(180° - 15°) = - cos 15° = -

4
V6-V2
sin 165° = sin(180° - 15°) = sin 15°
Therefore, the vertices of the hexagon that are in the second quadrant are
/— /— / j— — V6 + V2 rr V6 + V2
(3 V2 cos 105°, 3 V2 sin 105°) = 3V2---,3V2-
and
r- r ( r -V6-V20 f- V6- V2\

(3 V2 cos 165°, 3 V2 sin 165°) = 3V2---,3V2--- =
89
5.3.1
(a) In two dimensions, the point with rectangular coordinates (-2 V2,2 V2) has polar coordinates (4, . There¬
fore, the cylindrical coordinates of the point with rectangular coordinates (-2 ,
V2,2 V2 4) are
Turning to spherical coordinates, we have p - \J(-2 a/2)2 + (2 V2)2 + 42 = 4 V2. The 0-coordinate is
the same in spherical coordinates as in cylindrical coordinates. To find the (^-coordinate, we note that
z = p cos cp, so 4 = 4 V2cos cp. Therefore, we have cos cp = which means cp - |. The desired spherical
coordinates then are (4V2,f,j)
(b) In two dimensions, the point with rectangular coordinates (3,0) has polar coordinates (3,0). Therefore, the
point in space with rectangular coordinates (3,0, -3 V3) has cylindrical coordinates (3,0, -3 V3)
Turning to spherical coordinates, we have p - (3)2 + (0)2 + (-3 V3)2 = 6. The 0-coordinate is the same
in spherical and cylindrical coordinates, so 0 = 0. Finally, from z - p cos cp, we have -3 V3 = 6 cos cp, from
which we have coscp = -yr- This gives us cp - y, and the desired spherical coordinates are
5.3.2 Since x2 + y2 - r2, we have p = ^Jx1 + y2 + z2 = Vi/r2 + z2
5.3.3 We first tackle the case in which j < cp < tl. Let P be the point with z
spherical coordinates (p, 0, cp) and rectangular coordinates (x, y,z). Let O be
the origin and let B be the foot of the altitude from P to the xy-plane. We
have ZPOB = cp - |, so
PB = OP sin LPOB - p sin (^ _ ^) - ~P cos cp.
Since P is below the xy-plane, the z-coordinate of P is -PB - p cos <fi. We

also have
OB = pcos ZPOB = pcos (0 - —) = psirup.
y
Therefore, the x- and y-coordinates of B are x = p sin (pcos 0 and y =
p sin <p sin 0. The x- and y-coordinates of P are the same as those of P,
so the rectangular coordinates of P are (p sin <p cos 0, p sin tp sin 0, p cos cp), as
desired.
If cp - then (p sin (p cos 0, p sin <p sin 0, p cos (p) becomes (p cos 0, p sin 0,0), which are indeed the correct
coordinates. (To see why, recall that if (r, 0) are the polar coordinates of a point in the Cartesian plane, then the
rectangular coordinates of the point are (rcos 0,rsin 0).)
If <p ~ 7i, then P is on the z-axis, and the point is p below the origin. When (p = n, the rectangular coordinates
(p sin cp cos 6, p sirup sin 0, p cos <p) are (0,0, -p), so our expressions relating rectangular to spherical coordinates
work when cp - n.
Finally, if cp = 0, then P is on the z-axis, and the point is p above the origin. When cp - 0, the rectangular
coordinates (p sin cp cos 0, p sin cp sin 0, p cos cp) are (0,0, p), so our expressions relating rectangular to spherical
coordinates work when cp - 0.
90
Section 5.3
5.3.4 We have
V'(x-0)2 + (y-0)2 + (z-0)2 = \J (p cos 0 sin <p)2 + (p sin 0 sin cp)2 + (p cos cJS)2
= yjp2 cos2 0 sin2 <p + p2 sin2 0 sin2 <p + p2 cos2 <p

= \Jp2{cos2 0 + sin2 0) sin2 <ft + p2 cos2 (p
= a/p2 sin2 <p + p2 cos2 (p = \/p2(sin2 <p + cos2 <p) = =
5.3.5 For each part, let O be the origin and P' be the image of the described transformation.
(a) If P is on the xy-plane, then P' is the same as P. Otherwise, the xy-plane is between P and P'. Since P' is
the reflection of P through the xy-plane, the midpoint of PP' is on the xy-plane, and PP' is perpendicular
to the xy-plane. Let A be intersection of PP' and the xy-plane, so A is the foot of the perpendiculars from
P and from P' to the xy-plane. The polar coordinates of A in the xy-plane give us the r and 6 coordinates
of both P and P' in cylindrical coordinates. Therefore, the r- and 0-coordinates of P' are the same as those
of P. However, because P and P’ are on opposite sides of xy-plane, and are equidistant from the xy-plane,
their z-coordinates are opposites. Therefore, the cylindrical coordinates of P' are (r, 0, -z) . Note that these
also hold when P is on the xy-plane, since then we have z = -z = 0.
(b) Rotation about the z-axis doesn't change the distance from the point to the xy-plane, nor does it change the
distance between the origin and the foot of the altitude from the point to the xy-plane. Therefore, the r-
and z-coordinates of P' are the same as those of P. The only change is in the 0-coordinate. The foot of the
altitude from P' to the xy-plane is a rotation of tl around the origin of the foot of the altitude from P to the
xy-plane. Therefore, the cylindrical coordinates of P' are (r, 0 + n, cj))
(c) The reflection of P through the origin is the same as rotating P by n about the z-axis, and then reflecting the
result through the xy-plane. Therefore, P' is the result of performing both of the transformations described
in parts (a) and (b), which means its cylindrical coordinates are (r, 9 + n, -z)
We also could have reasoned as follows. Let A and A' be the feet of the altitudes from P and P',
respectively, to the xy-plane. Then, triangles PAO and P'A'O are congruent. Since PA = PA', but P and P'
are on opposite sides of the xy-plane, the z-coordinates of P and P' are opposites. Because AO = A'O, the
r-coordinates of P and P' are the same. Because the rays from O to A and A' are in opposite directions, the
0-coordinates of P and P' differ by n.
5.3.6 Yes. | The graph of (p - 0 and the graph of <p = n are both the z-axis, and the graph of <p | is the xy-plane.
This is why lines and planes are sometimes considered "degenerate" cones.
5.3.7 We tackle parts (a) and (b) together. The r-coordinate is the distance from the point to the Earth's axis, and
the z-coordinate is the distance from the point to the plane containing the equator. So, traveling with constant
latitude holds both the r-coordinate and the z-coordinate constant. Since the 0-coordinate is the same in cylindrical
coordinates as in spherical coordinates, traveling with constant longitude holds the 0-coordinate constant (while
varying r and z). So, matters are a little better in cylindrical coordinates than in rectangular coordinates—we
can describe east-west travel as varying just one coordinate. But they aren't as convenient as with with spherical
coordinates, since north-south travel still requires varying two coordinates. (In summary, the answer to every
question in this problem is "yes.")
91
5.3.8
(a) The graph of p = 4 is a sphere centered at the origin with radius 4. The
graph of 9 - np in spherical coordinates is the same as the graph of
9 = y in cylindrical coordinates. In the text, we explained why the
graph of 9 = c in cylindrical coordinates, where c is a constant, is a plane
through the origin. Therefore, the points such that p = 4 and 9 = are
the intersection of a sphere centered at the origin and a plane that passes
through the origin. This intersection is a circle with radius 4 centered
at the origin such that the circle is in the plane described by 9 —
= 2ti
Another way to see this is to note that the desired graph consists of
all points of the form (4, ^f, cp). When cp - 0, we have the point on the
z-axis, 4 units above the origin. As we increase cp from 0 to |, we trace
out a quarter circle centered at the origin from (0,0,4) (in rectangular
coordinates) to a point in the xy-plane that is 4 units from the origin.
Continuing to increase cp from | to n, we trace out another quarter circle
from the point in the xy-plane to (0,0, -4). As we increase cp from n to
2rc, we trace out the remainder of the circle.
(b) When 9 = j, we have x - pcos0sinc/> = 0, y = p sin 9 sin cp = psincp, and z = pcoscp. Therefore,
the entire graph is in the yz-plane. Letting p = 6 sin cp in our equations for y and z gives y = 6 sin2 cp
and z = 6 cos (p sin <p. The double angle identities for sine and cosine give us sin 2<p = 2 cos <p sin cp and
cos 2<p = 1 - 2 sin2 cp, so we can write y and z as y = 3-3 cos 2cp, z = 3 sin 2cp. We can now eliminate cp with
cos2 2(p + sin2 2(p — 1. We have
cos2 2<p + sin2 2<p = ( ) + (|) ,
so cos2 2<p + sin2 2<p = 1 gives us (3 - y)2 + z2 = 9. The graph of this equation is a circle in the yz-plane with
center (0,3,0) and radius 3.
Another way to approach this part is to note that when 9 = |, we have x = 0, y = p sin <p, and z = p cos (p,
so p2 = y2 + z2. Multiplying p = 6 sin cp by p gives p2 = 6p sirup, so y2 + z2 = 6y. Completing the square
leads to (y - 3)2 + z2 = 9. The graph of this equation is a circle in the yz-plane with center (0,3,0) and radius
3.
5.3.9
(a) We can use the conversion from spherical to rectangular coordinates to generate our parameterization. The
graph of the equation is a sphere with radius 2 and center (-3, -5,1). A sphere with radius 2 centered at the
origin has parameterization x = p cos 9 sin <p - 2 cos 9 sin <p, y = p sin 9 sin cp - 2 sin 9 sin cp, z = p cos cp =
2 cos cp. If we add -3 to the x-coordinate, -5 to the y-coordinate, and 1 to the z-coordinate of every point on
this sphere, we get the sphere with radius 2 and center (-3, -5,1). Therefore, the desired parameterization
is
x = -3 + 2 cos 9 sin cp, y - -5 + 2 sin 0 sin cp, z = 1 + 2 cos cp .
(b) We must restrict our choices of 9 and cp such that z > 0. Since z doesn't depend on 9, we don't restrict 9.
However, z > 0 gives us 1 + 2 cos cp > 0, from which we have coscp > -\- Since cos = ~\, the inequality
cos cp > tells us that we must restrict cp to 0 < cp < Therefore, the desired parameterization is
2n
x = -3 + 2 cos 9 sin cp, y = -5 + 2 sin 9 sin <p,z = 1+2 cos cp, 0 < cp <
T
92
Review Problems
Review Problems
5.25 Solving* = 3a-l fora givesa = (x+l)/3. Substituting into y = 2 a2-a

gives
x+1 X + 1 2x2 + 4x + 2 x+1 2x2 + x - 1

y=2
The graph of this equation is a parabola, as shown at right. Since * can take
on any value, the full parabola is produced by graphing the parametric
equations.
5.26 Completing the square in * and y gives (x + 2)2 + (y - l)2 = 16.

The graph of this equation is a circle centered at (-2,1) with radius 4. The unit circle can be parameterized by
x = cos 0,y - sin 0, so the circle with radius 4 centered at the origin can be parameterized by x = 4 cos 0, y = 4 sin 0.
Translating this circle 2 units left and 1 unit up produces the circle that is the graph of (x + 2)2 + (y - l)2 = 16, so a
parameterization of this circle is x = -2 + 4 cos 0, y = 1 + 4 sin 0
Another way we could have found this parameterization is by noting that cos2 0 + sin2 0 = 1, so we have
(4 cos 0)2 + (4 sin 0)2 = 16. Therefore, letting x + 2 = 4 cos 0 and y - 1 = 4 sin 0 gives (x + 2)2 + (y -1)2 = 16. Solving
these equations for x and y gives the parameterization x = -2 + 4 cos 0, y = 1 + 4 sin 0.
5.27 Applying the sine double angle identity, we have x = \ sin 21. Therefore, we have
x2 = t sin2 2f = 1(1 - cos2 2f) = \ (l - |) .
Solving for y gives y = 2 - 8x2. The graph of this equation is a parabola. However,
from x- \ sin2f, we see that x is restricted to the interval \—\, , and that all values of
x in this interval are attained. Therefore, the graph of these parametric equations is the
portion of the parabola within this interval, as shown at right.
5.28 The right side of x = cannot equal zero, so x is nonzero and we can divide
y = by x = jzp to eliminate the denominators. This gives us \ - t. Substituting
this into the given equation for x gives x = = jfryi- Since we know that x is
nonzero, we can divide both ends by x, and then multiply by x2 - y2 to give x2 - y2 = x.
Rearranging and completing the square gives (x - - y1 - The graph of this
equation is a hyperbola.
We know that all points in the graph of the parametric equations are on this hyperbola,
but we still must check if every point on the hyperbola is in the graph of the parametric
equations. First, we note that we must omit the origin from the graph of the parametric
equations, since x cannot be 0 in x = jzp. Are there any other points we must omit?
We cannot have 0 < x < 1, since 1 - f2 cannot be greater than 1. However, there are no points on the graph of
(x — |)2 - y2 = \ with 0 < x < 1, since y2 = (x - \) - \ < 0 for 0 < x < 1. Solving x = jzp for t gives t = ± yj 1 -
For x < 0 or x > 1, we have 1 - \ > 0, so for all values of x except 0 < x < 1, we can find a value of t such that
x = pL. Except when t - 0 (so x = 1), there are two such values of t, one negative and one positive. Since y = jzp,
the positive value of t produces the corresponding point on the hyperbola above the x-axis, and the negative value
of t produces the point below the x-axis. The graph is shown at right above; it consists of the entire hyperbola
except the origin.
93
5.29
(a) We have x = — 7 and y = -7, so r - \Jx2 + y2 = (-7)2 + (-7)2 = V98 - 7 V2. We seek an angle 0 such that
7 V2cos 0 = —7 and 7 V2 sin 0 - -7, which means cos 0 = -4^ arid sin 0 = We can take 0 = ^ ~ 5.50.
Hence, the polar coordinates are (7 V2,5.50) , where the angle is in radians.
(b) We have x = 6 V3 and y = -6 V2, so r = y^x2 + y2 = y/(6 V3)2 + (-6 y/2)2 = Vl80 = 6 V5. We seek an angle
0 such that 6 V5 cos 0 = 6 V3 and 6 V5 sin 0 = —6 V2, which means cos 0 = and sin 0 = - y/§. Since the
point (6 V3, -6 V2) lies in the fourth quadrant, we can take 0 = arcsin (- y^) ~ -0.68, which is equivalent
to -0.68 + 2tt = 5.6 radians. Hence, the polar coordinates are (6 V5,5.60) , where the angle is in radians.
(The answer (6 V5, -0.68) is also acceptable.)
(c) We have x = —5 and y = 8, so r = yjx2 + y2 = (-5)2 + 82 = V89- We seek an angle 0 such that
V89 cos 0 = -5 and V89 sin 0 = 8, which means cos 0 = - ^P and sin 0 = . Since the point (-5,8) lies
in the second quadrant, we can take 0 = arccos (-^g(p) ~ 2.13 radians. Hence, the polar coordinates are
(V89,2.13) , where the angle is in radians.
5.30
(a) We have r = 5 and 0 = 0°, so x = 5 cos 0° = 5 • 1 = 5, and y = 5 sin 0° = 5 • 0 = 0. Hence, the rectangular
coordinates are (5,0)
(b) We have r - — 4 and 0 = 210°, so x = -4cos210° = -4 • (-4^) = 2 V3, and y = -4sin210° = -4 • - 2.
Hence, the rectangular coordinates are (2 V3,2)
(c) We have r - 8 and 0 = -j, so x = 8cos (-f) = 8 • 4jp = 4 V2, and y = 8sin (~f) = 8 • (---r) = -4 V2.
Hence, the rectangular coordinates are (4 V2, -4 V2)
5.31 From the relations r2 = x2 + y2, x = r cos 0, and y = r sin 0, the given equation becomes x2 + y2 = 3x - 5y - 2.
Moving all the variable terms to one side, we get x2 - 3x + y2 + 5y = -2. Completing the square in x and y, we get
(x- |) + (y+§) = ^2- Hence, the graph is a circle with center (|, — |) and radius
5.32
(a) In the text, we showed that the point that results from rotating (x, y) counterclockwise by the angle 0 about
the origin is the point (x cos 0 — y sin 0, y cos 0 + x sin 0). A 60° clockwise rotation is a 300° counterclockwise
rotation, so the point that results when the (4,5) is rotated 60° clockwise about the origin is
(4 cos 300° - 5 sin 300°, 5 cos 300° + 4 sin 300°) =
94
Review Problems
(b) Let P be the point that results when (4,5) is rotated 60° clockwise about (2,5). We know
how to handle rotations about the origin. Since point (4,5) is 2 units to the right of (2,5),
we consider the 60° clockwise rotation of point (2,0) about the origin (because (2,0) is
2 units to the right of the origin). Let the result of this rotation be P'. Since (2,0) has
the same relationship to the origin that (4,5) has to (2,5), point F will have the same
relationship to the origin that P has to (2,5).
Applying the same formula we used in the previous part, we find that point P' is
(2 cos 300° - 0 sin 300°, 0 cos 300° + 2 sin 300°) = (1, - V3).
Since P' is 1 to the right of and V3 below the origin, point P is 1 to the right of and V3 below (2,5), which
means its coordinates are (3,5-V3)
(c) We apply the same approach as in the previous parts. Let P be the desired point. Since (4,5) is 6 units to the
right of and 2 units below (—2,7), we let point P' be the result of the 60° clockwise rotation of (6, —2) about
the origin. Therefore, point P' has coordinates
(6 cos 300° - (-2) sin 300°, (-2) cos 300° + 6 sin 300°) = (3 - V3, -1 - 3 V3).
So, point P is (-2 + (3 - V3), 7 + (-1 (1-V3,6-3V3)
5.33
(a) It is not always possible. For example, take r\ - = 1, 0\ - 0, and 02 = 180°. Then (ri,0i) = (1,0)
and (r2,62) - (1/180°) are the polar coordinates of the points with rectangular coordinates (1,0) and (-1,0),
respectively. Their midpoint is the origin, which has magnitude 0. But any point of the form (^y1,0) = (1,0)
always has magnitude 1.
(b) It is not always possible. For example, take r\ = 2, r2 = 1, 0i = 0, and 02 = 180°. Then (ri, 0i) = (2,0)
and (r2, @2) = (1,180°) are the polar coordinates of the points with rectangular coordinates (2,0) and (-1,0),
respectively. Their midpoint is the point ( \, 0), which lies on the x-axis but not the y-axis. But any point of
the form (r, ) = (r, 90°) lies on the y-axis.
5.34
(a) Let (r cos 0, r sin 0) be a point in the graph of f{r, 0) = 0. Since f(-r, 0) = /(r, 0), we have f(-r, 0) = 0,
so the point (-r cos 0,-r sin 0) also lies on the graph. The point (-r cos 0, -r sin 0) is a 180° rotation of
(r cos 0, r sin 0) about the origin is also in the graph. Therefore, the graph has 180° rotational symmetry
about the origin. (In other words, for any point P in the graph, the 180° rotation of P about the origin is also
in the graph.)
(b) Let (r cos 0, r sin 0) be a point on the graph of /(r, 0) = 0. Since /(r, 0) = /(r,-0), we have /(r,-0) = 0, so
the point (rcos(-0),rsin(-0)) = (rcos 0,-r sin 0) also lies on the graph. The point (rcos0,-rsin0) is the
reflection of the point (r cos 0, r sin 0) over the x-axis. Hence, the graph must be symmetric about the x-axis
5.35
(a) We have r = \Jx2 + y2 = 2 V2. From x = rcos0 and y = rsin0, we have cos0 = ^ and sin0 = so
0 = Therefore, cylindrical coordinates of the point are (2 V2, ^,2 V2)
We have p = yjx2 + y2 + z2 = 4, and the 0-coordinate in spherical coordinates is the same as the 0-
a/?
coordinate in cylindrical coordinates. Finally, from z = pcoscj), we have cos (p = -p = -5-, so cp = |.
Therefore, spherical coordinates of the point are
95
(b) We have r - \Jx2 + y2 = 2 V3. From x = r cos 0 and y - r sin 0, we have cos 0 = \ and sin 6 = 4^, so 0 = f.
Therefore, cylindrical coordinates of the point are (2 V3, |, -2 V3j
We have p = \Jx2 + y2 + z2 = 2 a/6, and the 0-coordinate in spherical coordinates is the same as the
Z — _YI ~~ A — 371
0-coordinate in cylindrical coordinates. Finally, from z = p cos <p, we have cos <p — ~Y, so (p 4 •
Therefore, spherical coordinates of the point are (2V6,f,f)
5.36
(a) We have x = r cos 0 = 6 cos y- = 6 = -3 and y = r sin 0 = 6 sin y = 6 (y) = 3 V3, so the rectangular
coordinates are (-3,3 V3, -1)
(b) We have
71 371 V2\ / V2'

x = p cos 0 sin (h = 12 cos — sin — = 12 ——
p r 4 4 \ 2
= 6,
■ n ■ , ir, ■ 71 . 3tz
y - p sm 0 sm<p = 12 sin — sm -—=12 1
/ V2s '^=6,
V2'
z = p cos <ft = 12 cos “ = 12 = -6V2,
so the rectangular coordinates are (6,6, -6 V2)
5.37 The portion of the graph of y2 + z2 = 9 that lies in the plane x = c is a circle in the plane with radius 3
centered at the (c, 0,0). This is true for any value of c. So, every cross-section perpendicular to the x-axis is a circle
with radius 3 and with center on the x-axis. Therefore, the graph is a
cylinder with radius 3 and the x-axis as its axis
5.38 We have r = 6 - z. So, for any particular value of z, the value of r is fixed, and z
0 ranges from 0 to 27Z. Therefore, each cross-section of the graph perpendicular to the
z-axis is a circle centered at the point on the cross-section with x = y = 0, except when
z = 6, for which the cross-section is the point (0,0,6). The radius of the cross-section
is directly proportional to the distance between the cross-section and the plane z = 6,
so the graph is a pair of cones as shown at right. The cones have vertex (0,0,6), and
the z-axis is the axis of the cones.
5.39
(a) The value of r in cylindrical coordinates is the distance from the point to the z-axis (if r > 0). Therefore, any
point on the graph of r = 5 is 5 units from the z-axis, so the graph is a
cylinder with radius 5 and with the z-axis as its axis .
(b) The equation p = 3 describes all points that are 3 units from the origin. Therefore, the graph is a
sphere centered at the origin with radius 3 .
(c) In polar coordinates, the graph of 0 = -| is a line through the origin that makes an angle of | with the
x-axis such that the line passes through the second and fourth quadrants. Since we can use any value for the
96
Review Problems
z-coordinate in the graph of 9 = in three dimensions, the graph of 9 = consists of the aforementioned
line in the plane z = 0, as well as every point "above" or "below" (i.e., with any z-coordinate) the line in
space. In other words, the graph is a plane that contains the z-axis and is perpendicular to the xy-plane.
(d) Let O be the origin and P be a point that has p = ^ in spherical coordinates. The ^-coordinate gives us the
angle that OP makes with the positive portion of the z-axis. Therefore, any point on the ray OP has p = ^
when expressed in spherical coordinates. Moreover, we can rotate any point on this ray about the z-axis
by any angle to produce another point with the same (^-coordinate. In other words, we can form a portion
of the graph by rotating the whole ray around the z-axis, thereby sweeping out a cone. Since we can allow
p to be negative, the graph consists of two cones with the z-axis as the axis and the origin as the vertex.
One cone is below the xy-plane (for p > 0) and the other is above the xy-plane (for p < 0).
5.40 Since r2 = x2 + y2, we can write the equation as x2 + y2 + z2 = 2z. Rearranging and completing the square in
z gives x2 + y2 + (z - l)2 = 1. Therefore, the graph consists of all points that are 1 unit from (0,0,1), which means
the graph is the sphere with center (0,0,1) and radius 1
5.41 First, we note that the origin is in the graph, since <p = \ gives p = 0. Multiplying both sides of p = 6 cos p by
p gives p2 = 6pcos(£, so we have x2 + y2+z2 = 6z. Rearranging and completing the square gives x2+y2 + (z-3)2 = 9,
so the graph is a sphere with radius 3 centered at (0,0,3).
5.42 We present two solutions.
Solution 1: Reason geometrically. Let P be a point on the graph, A be the foot of the altitude from P to the z-axis,
and O be the origin. Suppose we have 0 = so PA = p sirup. The equation
p = —^ gives p sin p - 3, so PA = 3. This means that P is 3 units from the z-axis. Conversely, any point with
0 < p < | that is 3 units from the z-axis satisfies p sin p = 3, so is on the graph.
If p = |, then sin p - 1, and the graph is the circle in the xy-plane with radius 3 centered at the origin.
If | < p < 7i, then LPOA - n — p. However, we still have sin(/) = sin(7i - p) = sin ZPOA = ^ = r-~ as before.
So, all the points that are 3 units from the z-axis and that have f < p < tl are also the graph.
Since sin p = 0 for p = 0 and for p = n, no points with (^-coordinate 0 or n are on the graph. Combining
all these observations, we see that the graph consists of all points that are 3 units from the z-axis. Therefore, the
graph is a cylinder with radius 3 and the z-axis as its axis.
Solution 2: Algebra. That the graph is a cylinder suggests a connection to cylindrical coordinates. From
p = we have p sin p - 3. Therefore, we have x = p cos 9 sin p = 3 cos 6 and y = p sin 6 sin p - 3 sin 9. Alia!
Considering the equations x = 3 cos 9, y - 3 sin 0 in cylindrical coordinates gives r = y^x2 + y2 = 3. We have no
restriction on z, so the graph is the same as the graph of r - 3 in cylindrical coordinates, and again we see that the
graph is a cylinder with radius 3 and the z-axis as its axis.
97
5.43
(a) As indicated in the diagram, adding n to the 0-coordinate rotates
the point by n about the z-axis. The 0-coordinate does not affect
the distance from the point to the origin (p), or the distance from
the point to the z-axis (r in cylindrical coordinates), or the dis¬
tance from the point to the xy-plane (z in rectangular coordinates).
Therefore, the transformation is a rotation of the point by n about
the z-axis.
(b) Since sin(n - <ft) = sine/), the x- and y-coordinates of the point
are unaffected. Since cos(7i - <p) = - cos <p, the z-coordinate of
the point with spherical coordinates (p, 0, (p) is the opposite of the
z-coordinate of the point with spherical coordinates (p, 6,tc - <p).
Therefore, the transformation is a reflection through the xy-plane.
Challenge Problems
5.44
(a) The graph of r = sin 20 is shown below. To see how this is the shape formed, consider first the values
0 < 0 < f. As 0 increases from 0 to the value of r goes from 0 to 1 and then back to 0, and we form the
petal in the first quadrant. Then, as 0 goes from 5 to n, the value of r goes from 0 to -1 and back to 0. This
forms the petal in the fourth quadrant, not the second, because r is negative. As 0 goes from 7i to =y, r goes
from 0 to 1 and back to 0, forming the third quadrant petal. As 0 goes from A1 to 2n, r again goes from 0 to
-1 and back to 0, forming the second quadrant petal, because r is negative.
(b) The graph of r - sin 30 is shown above. Here, the first quadrant petal corresponds to 0 < 0 < The
downward pointing petal corresponds to | < 0 < since when 0 is in this range, the value of 30 is
between n and 2tl, which means r - sin 30 is negative. The second quadrant petal is formed for < 0 < n.
When 7i < 0 < 31, the value of 30 is between ?>n and 471, which means r is negative. Therefore, the
points produced by these values of 0 are in the first quadrant, not the third. Moreover, each of the points
has already been produced when 0-0-f’ For example, when 0 = 7-f, we have sin 30 = -1, and the point
with polar coordinates (-1, is the same as the point with polar coordinates (l, which is produced
when 0 = |.
Similarly, when ^ < 0 < ^, we reproduce the downward pointing petal, and when ^ < 0 < In, we
reproduce the second quadrant petal.
98
Challenge Problems
(c) The graph of r = sin 40 is shown below.
(d) The graph of r - sin 50 is shown at right above.

(e) The graph of an equation of the form r = sinkd is sometimes called a "rose." We claim that if k is even,
then we obtain a graph with 2k "petals," and if k is odd, then the graph has k petals. To see why, we start
by looking at the graph of y = sin 2x for x e [0,27i], which is shown below.
The crest (a portion of the graph above the x-axis) on the interval [0, |] corresponds to the petal in the
first quadrant in the graph of r = sin 20. The trough (a portion of the graph below the x-axis) on the interval
[^,7i] corresponds to the petal in the fourth quadrant. (Even though an angle between | and n usually
indicates a point in the second quadrant, the petal lies in the fourth quadrant because r = sin 20 is negative
on this interval.) More generally, each crest and trough in the graph of y = sin2x (on the interval [0,27i])
corresponds to a different petal. There are four crests and troughs in the graph of y = sin 2x on the interval
[0,27t], hence there are four petals in the graph of r = sin 20.
In general, for even k, the high points of the crests occur at x = ||,..., , and the low points of
the troughs occur at x = ff, ..., which lead to the points (1, ^), (1, §f),..., (1, (4^3)TT), (-1, §f),
(—1, ^), ..., (-1, ^2^”) in polar coordinates. These 2k points form the "tips" of the petals. To show that
we obtain 2k petals, we must show that each petal tip is different.
Two petal tips can coincide if and only if they are of the form (1 ,<p) and (-1 ,<p + n). So, we must show
that the r-coordinates are not opposites when the 0-coordinates differ by n. If (1, cp) is on the graph, then
1 - sin k(p. When 6 = (p + n, we have
sin[k((p + 7z)] = sin (kcp + kn).
But k is even, which makes kn an even multiple of n. Hence, sin (kcp + kn) = sin kcp = 1. This tells us that
the point (1, (p + n) is also on the graph. So, for each petal in the graph of r - sin kd, there is a petal on the
opposite side of the origin. In particular, each petal tip is different, so the graph has 2k petals.
Now we look at the graph of y = sin 3x.
99
This graph exhibits 3 crests and 3 troughs, so we may expect 6 petals in the graph of r = sin 30.
However, the crest at x = p coincides with the trough at x = 7-f in polar coordinates, because p + n - -f,
sin(3 • f) = sin f = 1, and sin(3 • ?f) - sin y1 = -1. Hence, the 3 crests and 3 trough correspond to 3 petals
in the graph of r = sin 30.
For general odd k, we expect that each petal corresponding to a crest overlaps with another petal
corresponding to a trough. To show that this is the case, let (1, 0) be a point on the graph, so that 1 = sin/c0.
When 0 = 0 + 7i, we have sin[k(<p + rc)] = sin(kcp + kn). Since k is odd, kn is an odd multiple of n, so
sin(kd + kn) = -sin/c0 = -1. But the point (-1,0 + n) is the same point as (1,0). Hence, each petal
corresponding to a crest overlaps with another petal corresponding to a trough, so the graph of r = sin kd
has k petals.
5.45
(a) The graph of the parametric equations x = cos 0, y = sin 0 is the unit circle centered at the origin. Dilating
this circle by a factor of r, the graph of the parametric equations x - r cos 6, y = r sin 0 is the circle with
radius r centered at the origin. Translating this circle horizontally by h (adding h to the ^-coordinate) and
vertically by y (adding k to the y-coordinate) gives us the circle with radius r and center (h,k). Therefore,
parametric equations for this circle are x = h + r cos 0, y = k + r sin 0.
(b) The equations we use to convert spherical coordinates to rectangular coordinates give us parametric equa¬
tions for a sphere centered at the origin with radius p. These are x = p cos 0 sin 0, y = psin0sin0,
z = p cos 0. Adding a to the x-coordinate, b to the y-coordinate, and c to the z-coordinate of each point on
this sphere is equivalent to translating the sphere centered at the origin to the desired sphere. Therefore,
parametric equations for the sphere are x = a + p cos 0 sin 0, y = b + p sin 0 sin 0, z = c + p cos 0, where 0
and 0 are the parameters and p is a constant.
(c) Since x can be anything, we can let x - t be one of the parametric equations. For any t, the point (t, y, z)
must be on the circle centered at (t, 0,0) with radius r such that the circle is in the plane x-t. Therefore, we
must have (y - 0)2 + (z - 0)2 = r2, so y2 + z2 = r2. This means we can let y = r cos 0 and z = r sin 0. Parametric
equations for the cylinder then are x = f, y = rcos 0, z = r sin 0, where t and 0 are the parameters.
5.46 Completing the square in x and y gives (x - 2)2 + (y -1)2 = 4. The graph of this equation is a circle with center
(2,1) and radius 2. We can parameterize a circle with radius 2 as x = 2 cos t, y = 2 sin t, so we can parameterize the
circle with radius 2 and center (2,1) as x = 2 + 2 cos f, y = 1 + 2 sin t. However, graphing these equations produces
the whole circle; we only want the portion above the x-axis. We must only include those values of t for which
y > 0. From 1 + 2sint > 0, we have sint > Since sint < for t e yp], we can exclude the portion of
the circle below the x-axis and include the rest of the circle by limiting t to those values in the intervals [o, p1) and
(yp,2n]. We can write this as a single interval by noting that for t G oj, we produce the same values for x
and y as we do for t e (pp, 2n]. Therefore, we can parameterize the portion of the graph of x2 - 4x + y2 - 2y +1 = 0
that is above the x-axis with x = 2 + 2 cos t, y - 1 + 2 sin t, t e (-p, p1).
5.47 Let 9' = 2arctan ——^==. Because the range of arctangent is (~f, f ), we know that y is the unique angle
in the interval (-f, f ) that has tangent -——-. Therefore, 9' is the unique angle in the interval (—71,71) such that
tan v =-r=- Om‘ goal is to show that this angle is indeed the 0 such that (r, 0), with r > 0 and n < 9 < n, are
z x+-yx2+y2
the polar coordinates of the point (x, y).
We know that r - \Jx2 + y2, x = r cos 0, and y = r sin 0, so
y _ rsin0 _ sin0
x+ Jx2 + y2 ~ rcos9 + r ~ 1 +cos0'
100
Challenge Problems
By the half-angle formula for tangent, we have = tan |, so tan f = -7=. We showed above that only
2 x+ yx2+y2
one angle 9' in the interval (-n, n) satisfies tan ~ f- , so 9 must be this angle. Therefore 9 = 9' =
1 x+ yx2+y2
2 arctan-r= ■
x+ -yx2+y2
5.48 The distance between (1,2,3) and (2 - t, 4 + t, 3 + 21) is
y/(l - (2 - f))2 + (2 - (4 + f))2 + (3 - (3 + 2t))2 = y/(t - l)2 + (-2 - tf + (21)2 = V612 + 2t + 5.
Completing the square gives
V612 + 2t+ 5 =
Since 6 (f + g) is nonnegative for all real t, the minimum distance is , which occurs when t = - \
5.49
(a) Let C be the center of the circle, and let 9 be the angle that OX makes with the positive x-axis. Since XT is
parallel to the x-axis, we have Z.TXO = 9 as well. Therefore, we have ^ = cot 9. Since OT = 1, we have
TX = cot 9, so the x-coordinate of A is cot 9.
The y-coordinate of A matches that of point P. We have y
ZPOT = j — 9, and from isosceles triangle COP, we have
I.CPO = f - 9, so lOCP = n - ZPOC - zOPC = 29. We i T X
can find the y-coordinate of P by drawing an altitude from P to
the y-axis. Or, we can consider the parameterization of P as it
travels around the circle.
( c A
When 9 = 0, point P is at the origin, and when 9 = j, point P v ^P
is at T. For any value of 0 between 0 and we have ZOCP = 29. 0
One parameterization of a circle with radius \ centered at \ is
(\ cos f, \ + \ sin tj. Here, we want the parameterization in
terms of 9, the angle between XO and the x-axis. We have seen that lOCP = 29, so we let 29 be the
angle in the parameterization. We also require P to be at (0,0) when 9 = 0, so our parameterization is
Q cos (29 - |) , \ + \ sin (29 - f)). Applying the identity sin (x - f) = -cosx, the y-coordinate of P
then is \ - \ cos 29. The y-coordinate of A matches the y-coordinate of P, so parametric equations for A are
x = cot 9,y =\-\ cos 29 . As we'll see in the next part, we can also write this as x = cot 9, y = sin2 9.
(b) We have one trig identity that links cot 9 to sine or cosine, namely, cot2 9 + 1 = esc2 9. We can introduce a
sin2 9 to our parameterization x = cot 9, y = \ - \ cos 29 by using the double angle identity for cosine. We
have y =\~ \ cos 29 = \ -1(1 -2 sin2 9) = sin2 9. Substituting x = cot 9 and y = sin2 9 into cot2 9+1 = esc2 9
gives x2 + 1 = -, so the witch curve is the graph of
101
5.50 Parameterizing the red surface prior to the drilling is easy; we just borrow
from spherical coordinates to produce the parameterization x ~ 2 sin ft cos 6, y -
2 sin <ft sin 6, z = 2 cos ft. The hole complicates matters. If the hole were along the
z-axis, as shown at right, we could simply modify the parameterization we found for
the whole sphere. To see why let P be a point on the "top" of the drilled sphere, let
A be the foot of the altitude from P to the z-axis, and let O be the origin. We have
AP = 1 because the radius of the hole is 1, and right triangle OAP gives us OA - a/3.
Drilling the hole along the z-axis therefore has the effect of removing all points with
z-coordinate above V3 or less than -V3. We then must have — V3 < 2cosc/) < V3,
so -4^ < cos ft < -y, which means | < ft < y. This gives us the parameterization
x = 2 sin ft cos 6, y - 2 sin ft sin 6, z = 2 cos ft, where | < ft < y.
Turning to the surface formed when we drill along the x-axis, we see that we can't
simply limit ft or 0 alone in our initial parameterization for an un-drilled sphere.
However, we can reassign the parameterizations that we used for x, y, and z when
the hole was along the z-axis to produce a parameterization for the surface formed
when the hole is along the x-axis. We simply swap the parameterizations for x and z,
and we have
71
x = 2 cos ft, y 2 sin ft sin 6, z = 2 sin ft cos 8, where
6
5.51
(a) The red point travels in a circle around the center of B, but the center of B
moves! So, we first find parametric equations for the center of B. Let A be
the center of PA, point B be the center of B, and R be the red point. Point B is
always 2 units from A, so it traces out a circle of radius 2 as B rolls around
PA. We know how to parameterize a circle. Since the circle has radius 2,
we might think the parameterization is (2 cos t, 2 sin t). However, point B
starts at (0,2), so we can be more precise by using the parameterization
(2cos (f + f) ,2sin (ft + |H. Applying trig identities, we can write this
parameterization as (-2 sin t, 2 cos t).
Next, we must figure out where R is relative to B. Let X be the point
on PA at which PA and B are originally tangent. We draw vertical line BC
as shown through B, and we have ZCBA = iXAB = t because BC || AX. Let T be the point of tangency after
B has rotated about PA by angle t, as shown. Since B rolls along PA, the lengths of arcs RT and TX must be
the same. The radii of the circles are the same, which means that ZRBT = ZTAX = t, as well. Therefore,
we have ZCBR - 2t. So, as point B rotates counterclockwise by an angle t about point A, point R rotates
counterclockwise by 21 about point B. Since R is moving in a unit circle about the center of B, and starts
directly below the center of B, we can parameterize its position relative to B as cos (21 - to the right of B
and sin (21 - j'j above B. The "-j" accounts for the fact that R begins directly below B. Combining this
with the parameterization of B, the location of R is given by (-2 sin t + cos (2t - , 2 cos t + sin (21 — ).
Applying trig identities simplifies these coordinates to (-2 sin t + sin 21,2 cos t - cos 21)
(b) We saw in part (a) that point R is - cos 21 above point B. At t = 0, point R is -1 above point B (or, "1 below
point B"). As B is rolled around PA, the value of t ranges from 0 to 2n, so the value of 21 ranges from 0 to 471.
We have - cos 21 = -1 whenever cos2f = 1, which is whenever 21 is an integer multiple of 2n. Therefore,
as t ranges from 0 to 471, point R is directly below point B at t = 0, t - 2tc, and t = 471, so it makes [2] full
revolutions about the center of B. We also could have seen this by using the fact noted above that as B
102
Challenge Problems
moves an angle of t about A, point R moves 2f about B. So, as B goes around A once, point R goes around
B twice.
(c) The situation is only slightly different than in the first two parts, so we give
all the point labels the same meaning they had in part (a). We again find a
parameterization for B first. Point B moves in a circle of radius n + 1 about
point A and starts at (0, n + 1), so, by an argument similar to that in part (a),
we can parameterize point B with (-(« + 1) sin t, (n + 1) cos t).
The main difference comes in our calculation of iRBT. Arcs RT and XT

still have the same length, because 3 is being rolled along the circumference
of Ji. However, now the radii of the circles are different, so the angular
measures of the arcs are different. The length of an arc equals the radius
of the circle times the central angle that cuts off the arc. So, because arcs
RT and XT have the same length, we have (BT)(zRBT) = (AT)(/XAT). This
gives us LRBT - nt, so ZRBC = (n + 1 )t. Now, following the argument of part (a), the parameterization of R
is (-(« + 1) sin t + sin(n + 1 )f, (n + 1) cos t - cos(n + 1)0 , and point R makes n + 1 revolutions about B as
3 makes one full revolution about JK.
103
CHAPTER 6. BASICS OF COMPLEX NUMBERS
CHAPTER 6 i
l Basics of Complex Numbers
6.1.1 Isolating z2 gives z2 = -3, and "taking the square root of both sides" gives z = ± V—3 = ± V3 V—1 = ±/ V3
6.1.2
(a) a - b = (3 + 4f) - (12 - 5i) = 3 - 12 + 4z + 5i = | -9 + 9i \.
(b) We have ab = (3 + 40(12 - 5/) = 3(12 - 5i) + 4/(12 - 50 - 36 - 15* + 48/ - 20z'2 56 + 33/
(c) Factoring first makes this calculation easier:
a2 + 3a + 2 = (a + l)(a + 2) = (4 + 4z')(5 + 4z) = 20 + 16* + 20/ - 16 = 4 + 36/
a , « _ 3 + 4* i 3 — 4z 3 + 4/ 12 - 5i 3 - 4z 12 + 5 z
(d) +
T + ~b ~ 12 + 5z ^ 12-5/ 12 + 5/ 12 - 5z ' 12 - 5i 12 + 5/
(3 + 4z')(12 - 5z) (3 - 4z')(12 + 50 _ 56 + 33/ 56 - 33/ 112
“I” TTZ- _ <v _-~T — --TT- +
(12 + 5z)(12 - 5z') (12 - 5z')(12 + 5z') 169 169 169
Is it a coincidence that the result is real?
6.1.3
(a) 2/ + 7 - 2/ = —2/ + (7 + 2z‘) = [7].
1 1 3 + 5z 3 + 5z 3 5 .
(b) - -1
2 + 3z + 1 + 2z (2 - 3z') + (1 - 2z') 3 — 5/ 3 — 5/ 3 + 5z (3 - 5z')(3 + 5z') 34 34
6.1.4
(a) Since z2 = -1, we have z4 = (z2)2 = (-1)2 = 1 and z6 = (z2)3 = (-1)3 - —1. Therefore,
r,., /6 + /4 -1 + 1
0
^ - 7TT = TTT =
(b) Since (-z‘)2 = —1, we have (-z')4 = [(—02]2 = (~1)2 = 1, and (-z')6 = [(-z')2]3 = (-1)3 = —1. Therefore,
H)6 + (-04 -1 + 1
/(-0 = 0
—/ + 1 -/ + 1
104
Section 6.1
(c) First, we compute (z - l)2 = z2 - 2i +1 = -1 - 2i + 1 = -2z. Then (z - l)4 = [(z - l)2]2 = (-2i)2 = 4z'2 = -4, and
(z - l)6 = [(1 - z)2]3 = (-2z)3 = -8z3 = 8z. Therefore,
(z-1)6 + (z - l)4 8z -4 (8z -4)(-z)

f(i - 1) = 8 + 4z
(z - 1) + 1 z'(-z)
1,
6.1.5 Note that z 1 = 1/z = z'/z2 = z/(—1) = — z. Therefore, (z — z !) 1 = (z + z) 1 = =
2z 2z2 ~2Z
6.1.6
(a) Multiplying both sides of the equation by z - 3, we get z + 3z = 2z — 6. Subtracting z from both sides, and
adding 6 to both sides, we find that z = 6 + 3z
(b) Multiplying both sides of the equation by z, we get pp = (4 + 5z)z. Then dividing both sides by 4 + 5z, we
get
1 + 2z (1 + 2z)(4 - 5z) 14 + 3* 14 1 .
z = - “b -1
3(4 + 5z) ~ 3(4 + 5z)(4 - 5z) _ 3(42 + 52) 123 41
(c) Multiplying both sides of the equation by 1 + z, we get 3(1 + z')z + z = (10 - 4z')(l + z), which simplifies to
(4 + 3z)z = 14 + 6z. Dividing both sides by 4 + 3z, we get
14 + 6z _ (14 + 6z)(4 - 3z) _ 74 - 18* 74 18.

z=
4 + 3z _ (4 + 3z)(4 - 3z) “ 42 + 32 25 “ 25Z
6.1.7 Let iv = a + hi, Z\ = C\ + d\i, and Z2 = C2 + dzi. Then, by the definition of complex number addition, we have
Zi + Z2 = (ci + C2) + {d\ + rf2)*- By the definition of complex number multiplication, we have
w(z, 1 + Z2) = (zz + bi)((ci + C2) + (d\ + dî) = (zz)(ci + C2) — b{d\ + z^) (^(^1 £2) &{di d2))i-
We then apply the distributive property for real numbers to find
w(zi + Z2) = ac\ + ac2 - bdi - + (bc\ + bc2 + ad\ + ad2)i.
Next, we compute wz\ + wz2. By the definition of complex number multiplication, we have
wz 1 + WZ2 = ((zzci - Mi) + (Mi + &d)z) + {{ac2 - M2) + (M2 + bc2)i).
Then, applying the definition of complex number addition, followed by the associative property for real numbers,
we have
wz\ + WZ2 = ac\ + ac2 - bd\ - M2 + (Mi + M2 + Mi + ad2)i.
Therefore, we have w(z\ + Z2) = wz\ + wz2r so complex numbers are distributive.
6.1.8 We have i~n = (z-1)” = (|)”. Since 7 = 7 • f = ^ = -z, we have z'_” = (-z')w. Just as the powers of z repeat
every four terms, the powers of -z also repeat every four terms. So, we only have to check in + (-z')n for n = 0,1,
2, and 3, to determine all possible values of in + (-*)”. We have
z° + (-z')° = 2,
z + (-z)1 = z - z = 0,
z-2 + (—z)2 = -1 - 1 = -2,
z'3 + (—z')3 = -z + z = 0.
Hence, the possible values of S are -2,0, and 2
105
6.2.1 In the diagram at right, we see each of the points 4 + 7z, -6 - 2z, and Im
(3 + z')(-2 + 5z) = -11 + 13z plotted in the complex plane, similar to the way *+
the points (4,7), (-6,-2), and (-11,13) would be plotted in the Cartesian
coordinate plane.
:: 4 + 7/
6.2.2
Re
(a) The magnitude of 24 - 7z is y/24:2 + (-7)2 = V625 = 25
ilium ++ . ++
-6 - 2i
(b) The magnitude of 2 + 2 V3z is \fl2 + (2 V3)2 = Vl6 = _4
(c) The magnitude of (1 + 2z')(2 + i) = 5z is V52 = [5].
6.2.3 As we see in the diagram at right, the complex numbers w, z, w, and z

form an isosceles trapezoid. The lengths of the bases are \w - w\ - |10z| = 10
and |z - z| = |4z| = 4. The height is 12 - 3 = 9, so the area of the trapezoid is
o 10+4 _
y' 2
63
6.2.4 The distance between 4+7z and -3-17z is the magnitude of their difference:
2 = 12 - 2 z
|4 + 7i - (-3 - 17z)| = |7 + 24z| = V72 + 242 = V625 = 25
w - 3 - 5z
6.2.5 Let Zi = «i + b\i and z2= a2 + b2i. So, we have
Z\ + Z2 ZZ} + b\i + U2 + &2Z' 0-1 + Z?2 1--

b\ + &2
—-—. — --- = —-- — • l.
The midpoint of the segment connecting zi and Z2 on the complex plane corresponds to the midpoint of the segment
connecting (a\, bi) and (a2, b2) on the Cartesian plane. This midpoint on the Cartesian plane is (^Lr2, which
does indeed correspond to the point (zi + z2)l2 on the complex plane as shown by our calculation above.
6.2.6
1 + 2z _ (1 + 2z')(2 - 0 _ 4 + 3z _ 4 3 ..
(a) The magnitude of z is yj(4/5)2 + (3/5)2 = V25752 = 0.
2 + z ~ (2 + 0(2 - 0 5 ~ 5 + 5
(b) The magnitude of
6 + Hz' (6 + llz')(ll - 6z') _ 132 + 85* _ 132 85_
11 + 6z (ll + 6z')(ll-60 157 157 157
is v/(132/157)2 + (85/157)2 = V24649/1572 = 0.

(c) From parts (a) and (b), the magnitude of (x + yz’)/(y + xi) appears to be 1. We can prove this as follows. The
magnitude of
x + yi (x + yz‘)(y - xi) _ 2xy + (y2 - x2)i _ 2xy
_+
y + xi (y + xi)(y - xi) x2 + y2 x2 + y2 x2 + y2
is
2^y V2 - x2 x 2 4x2y2 ^ x4 - 2x2y2 + y4

x2 + y2 x2 + y2 (x2 + y2)2 (x2 + y2)2
/x4 + 2x2y2 + y4 _ /(x2 + y2)2

(x2 + y2)2 (x2 + y2)2 =0
106
Section 6.3
6.2.7 Every square is also a parallelogram, and the diagonals of a parallelogram

bisect each other. Plotting the points 1 + 2i, -2 + i, and -1 - 2/ in the complex plane,
we see that the numbers 1 + 2i and —1 — 2/ must lie at opposite corners of the square.
Hence, the midpoint of the diagonal between these points is [(1+2/) + (-1 -2z')]/2 =
0.
Re
Therefore, the midpoint of z and -2+i is also 0. In other words, [z+(-2+z)]/2 = 0,
so z = 2-/
There are many other ways we might have approached this problem; we could
have used slopes, distances, or even congruent triangles.
6.3.1 Instead of expanding the product, we can use the fact that the magnitude of the product is equal to the
product of the magnitudes:
|(10 + 24f)(8 - 601 = |10 + 24/1 • |8 - 6/| = VlO2 + 242 • ^82 + (-6)2 = V676 • VlOO = 26 • 10 = 260
6.3.2 Letz = a + bi, where a and b are real numbers. Then \z\ = \a + bi\ = Vfl2 + b2, and |z| = |a-bi\ = 'a2 + (—b)2 =
Vfl2" + b2. Hence, \z\ = |z|. We also could have noted that |z| = \J(z)(z) = V(z)(z) = M-
6.3.3 Let z = a + bi for some real numbers a and b. Then z = a - bi, and the given equation becomes
3(a + bi) + 4(a - bi) = 12- 5/ => 7a- bi = 12- 5i.
Two complex numbers are equal if and only if their real parts are equal and their imaginary parts are equal, so
12
we get 7a = 12 and b = 5. Therefore, z = a + bi = + 5/
y
We also could have solved for z without writing out the real and imaginary parts as follows. Taking the
conjugate both sides of the given equation 3z + 4z = 12 - 5/, we get 3z + 42 = 12 - 5/, so 3z + 4z = 12 + 5i. This gives
us the system of equations
3z + 4z = 12 - 5/,
4z + 3z = 12 + 5/.
Multiplying the first equation by 3 and the second equation by 4, we get
9z + 12z = 36 - 15/,
16z + 12z — 48 + 20/.
Subtracting the first equation from the second equation, we get 7z = 12 + 35/, so z = y + 5/.
6.3.4 Let z = a + bi, for some real numbers a and b.
(a) If z2 = 2/, then 2/ = (a + bi)2 = a2-b2 + 2 abi. Equating the real parts and equating the imaginary parts, we get
the system of equations a2 - b2 = 0 and 2ab = 2. From the second equation, we have b = 1/a. Substituting
this expression for b into the first equation, we get
a2 - \ = 0 => a4 = 1.
a2
107
Since a is real, we have a = ± 1. The corresponding values of b - 1/a are ±1. Therefore, the two square roots
of 2i are 1 + z and -1 - i
(b) If z2 — —5 + 12z, then ^5 + 12z - (a + bi)2 = a2 -b2 + 2obi. Equating the real parts and equating the imaginary
parts, we get the system of equations a2 -b2 = -5 and lab = 12. From the second equation, we have b = 6/a.
Substituting this expression for b into the first equation, we get
2 36 c
a-T = -5 a4 + 5a2 - 36 = 0 {a1 + 9)(a2 - 4) = 0.
a2
Since a is real, we have a = ±2. The corresponding values of b are 6/a = ±3. Therefore, the two square roots
of -5 + 12z are 2 + 3i and -2 - 3i
(c) If z2 = 24 - lOz, then 24 -10i = (a + bi)2 = a2 -b2 + labi. Equating the real parts and equating the imaginary
parts, we get the system of equations a2 - b2 = 24 and lab = -10. From the second equation, we have
b - -5/a. Substituting this expression for b into the first equation, we get
2 23
a2 -T = 24 z4 - 24fl2 - 25 = 0 {a2 - 25)(fl2 + 1) = 0.
a1
Since a is real, we have a = ±5. The corresponding values of b are b = —5/a - +1. (The + symbol indicates
that b = -1 goes with a = 5 and b - 1 goes with a = -5.) Therefore, the two square roots of 24 - lOz are
5 - i and -5 + i
6.3.5
(a) The magnitude of

1 1 3 - 4z 3 - 4z A _ A-
z 3 + 4z (3 + 4z)(3 - 4z) 25 25 25Z
is |l/z| = y/(3/25)2 + (4/25)2 = V25/252 = \l/5
(b) The magnitude of z is |z| -- V32 + 42 = V25 = 5, so l/|z| = 1/5
(c) First, we make the denominator of z/w real:
z _ 3 + 4z _ (3 + 4z')(5 + 120 _ (3 + 4z)(5 + 12z)

zv ~ 5 - 12z _ (5 - 12z)(5 + 120 _ 169 '
We leave the numerator as a product, because it makes the magnitude easier to compute:
1 1
(3 + 4z) • (5 - 120 |3 + 4z| • |5 - 12*|
169 169
= _L . • VpTiy = i . V25 ■ Vl69 = ^ . 5.13 =

13
(d) As computed in part (b), the magnitude of z is |z| = 5. The magnitude of w is\w\ = y^52 + (-12)2 = Vl69 = 13.
Therefore, \z\/\w\ = 5/13
It appears that \z/w\ = \z\/\w\. This holds for all complex numbers z and w, where w ? 0. This is because
z
\w\ = — ■ zv = |z|.
zv
Dividing both sides by \w\, we get M

N'
108
Section 6.4
6.3.6 Let z — a + bi, where a and b are real numbers. Then
_ + bi _ (a + bi)(a + bi) _ (a2 — b2) + 2abi a2 - b2 lab .

z a-bi (a- bi)(a + bi) a2 + b2 a2 + b2+ a2 + b2 Z'
(a) If z/z is real, then the imaginary part lab/(a2 + b2) must be 0. This is zero if and only if a = 0 or b = 0. Since
z = a + bi is real if and only if b = 0 and z is imaginary if and only if a = 0, we conclude that z/z is real if and
only if z is real or imaginary (and not equal to 0)
(b) If z/z is imaginary, then the real part (a2 - b2)/(a2 + b2) must be 0. This is zero if and only if a2 = b2, or a = ±b.
Therefore, z/z is imaginary if and only if z is of the form a ±ai — a( 1 ± i) for some nonzero real number a.
6.3.7 Let A: be a real number such that x = ki is an imaginary solution to the given equation. Substituting x = ki
into the left side gives
(fa')4 - 3(ki)3 + 5(fa)2 -17(ki) - 36 = k4 + 3k3i - 5k2 - 17ki - 36 = (k4 - 5k2 - 36) + (3A3 - 17k)i.
If this expression equals 0, then the real part and the imaginary part equal 0, so we have the system of equations
k4 - 5k2 - 36 - 0,
3/c3 - 17k = 0.
The first equation factors as (k2 - 9)(k2 + 4) = 0, so (k - 3)(k + 3)(k2 + 4) = 0. The second equation factors as
3k(k — 3)(k + 3) = 0. The common roots are k - 3 and k = -3. Therefore, the imaginary solutions of the given
equation are x = ±3i . (What are the other two roots?)
6.4.1 Note that all the coefficients are positive, so if there are any real roots, then they must be negative. Checking
negative integers that divide 18, we find g(—1) = 6 and g(—2) = -8, so there's a root between -1 and -2. If a/b is a
root of g(x) (in lowest terms), then by the Rational Root Theorem, a must divide 18 and b must divide 2. Our only
option between -1 and -2 is -3/2, so we try that. We find g(-3/2) = 0, so 2x + 3 is a factor of g(x) and -3/2 is a
root of g. Dividing by 2x + 3, we find 2x3 + 5x2 + 15x + 18 = (2x + 3)(x2 + x + 6). Applying the quadratic formula
to x2 + x + 6 gives us the roots x = (-1 ± i V23)/2, so the roots of g(x) are
3 -1 + z'V23 —1 — i V23
x = and
2 '
6.4.2 First, we look for integer roots. If there are any integer roots, then they must be among the factors of 36.
Checking these values, we find /(3) = /(-4) = 0, so (t - 3)(f + 4) is a factor of /(f). Dividing by (f - 3)(f + 4), we find
2f4 - 23f2 + 271 - 36 = (f - 3)(f + 4)(2f2 -21 + 3).
Therefore, the roots are t =
6.4.3 The coefficients of g are all real, so nonreal solutions are present in conjugate pairs. Since 2 + i and 3 + i are
roots of g, then so are 2 - i and 3 - i . That gives a total of 4 roots, which is the maximum possible, so there are
no other roots of g.
109
6.4.4 Since the nonreal roots of a polynomial with real coefficients come in conjugate pairs, the number of nonreal
roots of the polynomial is even. Since the polynomial is of degree 4, the total number of roots is 4, so the number
of real roots must also be even. We are given that the polynomial has at least one real root, so the polynomial must
have at least two real roots.
6.4.5 Let the integer roots be r and s. We can then write the polynomial in the form
(x - r)(x - s)(x2 + ax + b),
for some real numbers a and b. This expands as
x4 + (a - r - s)x3 + (b -ar-as + rs)x2 + (ars -br - bs)x + brs.
We know that all the coefficients are integers. The coefficient of x3 is a - r - s, and r and s are both integers, so
a is an integer. Then the coefficient of x2 is b - ar - as + rs, and ar, as, and rs are all integers, so b is also an integer.
Each of the roots listed is nonreal, so each must be a root of the quadratic x2 + ax + b, if it is to be a root at all.
Furthermore, the other root must be the conjugate. For example, if one root is 1+1 ^, then the other root must be
i-.o/n. gum q£ £hese roots is p and their product is
i + z'Vn l-zVn _ l + ii _0
2 2 “ 4 “ '
Hence, and lrare the roots of x2 - x + 3 = 0. This shows 1+1 ^ can be a root of the original polynomial,
so it must be the answer. We check the other choices to be sure:
Root Quadratic
l+i'VTT
2 x2 - x + 3
1+i
2 x2 - x + \
2 +i x2 - x + |
1 + i x2 - 2x + |
l+i Vl3
2 x2 - x + \
The only quadratic of the form x2 + ax + b where a and b are integers is x2 - x + 3, so the only possible root is
i+zVn
2
6.4.6 Note that all the coefficients are positive, so if there are any real roots, then they must be negative. First,
we look for integer roots (which must be negative factors of 5), and find that there are none. Since the quartic is
monic, all rational roots must be integers, so there are no rational roots either.
We know that the nonreal roots come in conjugate pairs. We also know that roots of quadratics with real
coefficients come in conjugate pairs. So, we hope that we can factor the quartic as the product of quadratics.
Maybe the quadratics will have integer coefficients that we can find. Let x4 + 5x2 + 4x + 5 = (x2 + ax + b)(x2 + cx + d),
where a, b, c, and d are integers. Expanding the right side, we get
x4 + 5x2 + 4x + 5 = x4 + (a + c)x3 + (b + d + ac)x2 + (ad + bc)x + bd.
110
Review Problems
Equating the corresponding coefficients on both sides, we obtain the system of equations
a + c = 0,
b + d + ac = 5,
ad + be - 4,
bd = 5.
Now, we have a number game much like factoring a quadratic. From the first equation, we have c = -a. The
equation bd = 5 suggests that we try the values b = 1 and d - 5. Then from the second equation, we have a = ±1
and c = +1. Only a = 1, c = -1 (together with b - 1, d - 5) satisfies the third equation, so we discard a = -1, c = 1.
Hence, we have the factorization
x4 + 5x2 + 4x + 5 = (x2 + x + l)(x2 - x + 5).
Therefore, the roots are x =
Review Problems
6.29
(a) 2w-3z = 2(2 + 3i) - 3(4 - 5i) = 4 + 6/ - 12 + 15/ -8 + 21/
(2-3z) 2-3 i 2 -3 i 2_ _ 3_.

(b) = 1
w 2 + 3i (2 + 30(2-30 22 + 32 13 13 13Z
(c) We begin by computing the denominator: w + z = (2 — 30 + (4 — 5z) = 6 - 8z. Then
2 _ 2 _ 1 3 + 4z 3 + 4z
_ 3 4 .
w+z 6 — 8z 3 — 4/ (3 — 4z’)(3 + 4z) 32 + 42 25 + 25Z
(d) First, we can factor the numerator and denominator:
iv3 + 2zv2z + wz2 _ w(w2 + 2wz + z2) _ w(w + z)2 _ w + z

w2z + wz2 wz(w + z) H7Z(ty + z) z
Then
w+z (2 + 30 + (4 - 50 _ 6 - 2z _ (6 - 2Q(4 + 5Q _ 34 + 22/ _ 34 22.
4-5/ 4 - 5z (4 - 50(4 + 5z') 41 41 + 41Z
(el = (~y + *0(*-y0 = = = [71. We could have also observed that

K ) x + yi (x + yi)(x - yz') x2 + y2 x2 + y2 LJ
-y + xz = z'(x + yz).
(f) Taking a cue from part (e), we can observe that 1 - i = -z(l + z‘). Therefore,
(l - o4 = (~04(1 + o4 _
1+z
(1 + if (1 + if
6.30 Grouping the x2 terms on one side and constants on the other gives 2x2 = -12. Dividing by 2 gives x2 = -6.
Taking the square root of both sides gives x = + V~6 = + V6 V~1 - ±/' V6
111
6.31 We evaluate the expression one step at a time, starting from the bottom. We have
1 - 1+ 1 _
i ~ *
1+/ 1+/ 1+2

but we don't bother rationalizing the denominator because we're going to take the reciprocal of this expression in
the next step:
1+2 2+1+2 1 + 22
1+ = 1 + = 1 +
i
1-
1 + 2 1+2
Then
1 _ 1 2 2 1-2/ 2 — 2 22 2 1.
~1 " (1 + 2i)/i ~ 1 + 22 “ 1 + 2/ ’ 1-2/ 5 5 + 5Z
6.32 Multiplying both sides by 1 + z, we get z = (-1 + /)(1 + z) = (-1 + /) + (-1 + /)z. Then (2 - i)z = -1 + /, so
-1 + / _ (-1 + /)(2 + /) _ -3 + / 3 1.
2-/ " (2 - /)(2 + /) “ 5 5 + 5l
6.33 Let z = a + bi, where a and b are real numbers. Then z = a - &/, which has the same
real part as z, but opposite imaginary part. Hence, z is the reflection of z over the real
axis. Similarly, —z -- -a + fo/ has the same imaginary part as z, but opposite real part.
Hence, —z is the reflection of z over the imaginary axis. Finally, -z = -a - bi has both
opposite real part and opposite imaginary part as z, so -z is the reflection of z through
the origin. We can then label -z, z, and —z as shown at the right.
6.34 We present two solutions.
Solution 1: The faster way.
(a) We showed in the text that a complex number is real if and only if it equals its conjugate. We have
wz + wz = wz + wz = w ■ z + w ■ z - wz + wz = wz + wz,
so wz + wz is real.
(b) Similarly, we showed that a complex number is imaginary if and only if it equals the negative of its
conjugate. We have
-(wz - wz) = -wz + wz = -w ■ z + w ■ z = -wz + wz - wz - wz,
so wz - wz is imaginary.
We could also have noted that the conjugate of wz is wz = w • z = wz. Hence, wz + wz is the sum of wz and its
conjugate, which we know to be real. (In particular, it is twice the real part of wz.)
Similarly, wz-wz is the difference between wz and its conjugate, which we know to be imaginary. (In particular,
it is 2/ times the imaginary part of wz.)
Solution 2: The slower way. Let w - a + bi and z = c + di, where a, b, c, and d are all real.
112
Review Problems
(a) In terms of a, b, c, and d, we have
wz + wz = (a + bi)(c - di) + (a- bi)(c + di)

= ac + bd + (-ad + bc)i + ac + bd + (ad - bc)i = lac + Ibd,
which is real.
(b) In terms of a, b, c, and d, we have
wz - wz - (a + bi)(c - di) -(a- bi)(c + di)

= ac + bd + (-ad + bc)i - (ac + bd) - (ad - bc)i - (-lad + lbc)i,
which is imaginary.
6.35 Let z = a + bi for some real numbers a and b. If z2 = 5 — 12i, then 5 — 12/ — (a + bi)2 = a2 -b2 + labi. Equating
the real parts and equating the imaginary parts, we get the system of equations a2 — b2 = 5 and lab = -12. From
the second equation, we have b = -6/a. Substituting this expression for b into the first equation, we get
36
= 5 fl4 -5a2- 36 = 0 (a2 - 9)(a2 + 4) = 0.
Since a is real, we have a - ±3. The corresponding values of b = -6/a are +2. (The + symbol indicates that b =
goes with a - 3 and b - 1 goes with a = —3.) Therefore, the square roots of 5 - 12/ are 3 - 2/ and -3 + 2
6.36 If 7 + / is 5 units from 10 + ci, then from the distance formula, we have
y/{7 - 10)2 + (1 - c)2 = 5 9 + (c - l)2 - 25 =» (c - if = 16 => c - 1 = ±4,
so c = 5 or -3
6.37 The value of \z — (4 - 5i)| is the distance between z and 4 - 5/ in the complex plane. The equation \z — (4 - 5z)| =
2 V3 tells us that this distance is 2 V3. Therefore, the graph of \z — (4 - 5/)| = 2 V3 is a circle centered at 4 - 5/ with
radius 2 V3, which has area (2 V3)2tz = 1271
6.38 Let z = a + bi be the complex number corresponding to the point F, where a and b are real numbers. From
the diagram, we can say two things about z. First, since F lies to the right of the imaginary axis and above the real
axis, a > 0 and b > 0. Second, F lies outside the unit circle, so |z| > 1. We must deduce similar facts about 1/z.
First we express \ as a complex number:
1 1 _ a - bi _ a - bi _ a b
z a + bi (a + bi)(a - bi) a2 + b2 a2 + b2 a2 + b21
Since a and b are positive, the real part of 1/z is positive, and the imaginary part is negative. Therefore, 1/z lies to
the right of the imaginary axis, and below the real axis. This means 1 /z must be A or C.
Next, we see that = — < 1. Hence, 1/z lies inside the unit circle. Therefore, the reciprocal of F is C
\z\
6.39 Let z-a + bi, where a and b are real numbers, so the equation becomes
\a + l + (b- l)i| = \a -1 + bi\

=* y/(a + l)2 + (b- l)2 = yj(a -1)2 + b2
=> (a + l)2 + (b - l)2 = (a- 2)2 + b2
=> a2 + la + 1 + b2 — lb + 1 = a2 — 4a + 4 + b2
=> 6a — lb + 1
=> b - 3a - 1.
113
Hence, all solutions are of the form z = a + (3a - l)z , where a is any real number.
Note that we could also have solved this problem by noting that |z + 1 - i\ can be written as \z - (-1 + i)\, so this
expression represents the distance from z to -1 + i on the complex plane. The expression \z-2\ equals the distance
from z to 2, so the solutions to the equation |z + 1 — z| = |z - 2| represent all z equidistant from —1 + z and 2 on the
complex plane. Therefore, z must be on the perpendicular bisector of the segment connecting — 1 + i and 2 on the
complex plane. This segment has slope and its midpoint is \ + So, z is on the line through \ + \ with slope
3. See if you can finish from here to get the same answer as above.
6.40 We are given that (a + bi)(c + di) - ac -bd+ (ad + bc)i is real, so ad + bc = 0. Then
(b + ai)(d + ci) - bd - ac + (be + ad)i = bd- ac + (ad + bc)i = bd - ac,
so the product of b + ai and d + ci is also real.
6.41
(a) Since (z - z)/(2i) equals Im(z) for all z, we divide the given equation by 2i to get (z - z)/(2z) = -4, which we
can write as Im(z) = -4. The graph is a horizontal line 4 units below the real axis, shown to the left below.
Re Re
(b) Let a = Re(z) and b = Im(z). Then the given equation becomes
(4 - i)(a + bi) - (4 + i)(a - bi) = 16z

=> 4 a + b - ai + 4bi -4 a-b- ai + kbi = 16z
=> -2ai + 8bi - 16i
=> -a + 4b - 8
b = - + 2.
4
Thus, the graph is a line with slope 1/4 that intersects the imaginary axis at 2i, shown to the right above.
(c) The left side can be rewritten as |7 + z - 2z| = \-(7 + i - 2z)| = |2z — 7 — i\, so |2z — 7 — z'| = 4. Dividing both
sides by 2, we get
7+z
z = 2.
This equation tells us that the distance from z to (7 + z)/2 on the complex plane is 2. Therefore, the graph is
a circle centered at \ + \i with radius 2.
114
Review Problems
6.42 For a complex number z, |z - 5 + 2i\ = |z — (5 - 2z)| is the distance Im

between z and 5 — 2z. Hence, the graph of |z — (5 — 2z)| < 4 is the set of all
complex numbers z that are within 4 units of 5 - 2i. This is the closed disk
(which is both the circle and its interior) centered at 5 - 2i with radius 4,
shown at right.
6.43 To deal with such large exponents, we compute the first few powers
of i + 1 to see if we can find a pattern:
(z + l)1 = z +1,
(z + l)2 = z2 + 2z + 1 = 2 z,
(z + l)3 = (z + l)2 • (z + 1) = 2i(i + 1) = -2 + 2z,
(z + l)4 = (z + l)3 • (z + 1) = (-2 + 2z)(z + 1) = -4.
We note that (1 + if is a real number, which makes it easy to work with. Similarly, we find that (z - l)4 = -4.
Therefore,
(z + l)3200 - (z - l)3200 = [(* + l)4]800 - [(z - l)4]800 = (-4)800 - (-4)800 =
We'll see an even faster way to solve this problem in the next chapter.
6.44 Let z - a + bi, for some real numbers a and b. We are given that |z| = 1, so |z| = |a + bi\ = Ja2 + b2 = 1, or
a2 + b2 - 1. Then
|z - 1|2 + |z + 1|2 = |(a - 1) + bi\2 + |{a + 1) + bi\2 = (a- l)2 + b2 + (a + l)2 + b2

— if — 2a + 1 + b2 + a2 + 2zz + 1 + b2 — 2 a2 + 2 b2 + 2
= 2 {a2 + b2) + 2 = 2 + 2 = 4.
Alternatively, we could have used the fact that zz = |z|2 for all complex numbers z:
|z - 1|2 + |z + 1|2 = (z - 1) • (z- 1) + (z + 1) • (z + 1) = (z- l)(z - 1) + (z + l)(z + 1)

-zz-z-z + 1 + zz + z + z + l = 2zz + 2 = 2\z\2 + 2 = 4.
Finally, can you find a geometric interpretation of the result?
6.45
(a) We have
|w - z\2 + Iw + z\2 = (w- z)(w -z) + (w + z)(zv + z)

= ww -WZ-ZW + ZZ + WW + WZ + ZW + ZZ
- 2ww + 2zz = 2(|w|2 + |z|2).
(b) Consider the parallelogram in the complex plane with vertices at 0, w, z, and w + z. The distance from
w to z is \zv - z|, so the expression |w - z\2 + \w + z|2 is the sum of the squares of the diagonals. The sides
of the parallelogram from the origin have lengths \w\ and |z|, and opposite sides of any parallelogram are
equal, so 2(|w|2 + |z|2) is the sum of the squares of the sides of the parallelogram. Therefore, the equation
2(M2 + |z|2) = |w- z\2 + |w + z\2 expresses the fact that the sum of the squares of the sides of a parallelogram
equals the sum of the squares of the diagonals.
6.46 Since the coefficients of the polynomial are real, the complex conjugate of -3 + 5z, namely -3 - 5z, must also
be a root. The sum of -3 + 5z and -3 - 5z is -6, and their product is (-3 + 5z)(-3 - 5z) = 9 + 25 = 34. Therefore,
115
the quadratic whose roots are -3 + 5z and -3 - 5z is x2 + 6x + 34. Since this quadratic must be a factor of the given
quartic, we have
x4 + 9x3 + 48x2 + 78x - 136 = (x2 + 6x + 34)(ax2 + bx + c)
for some constants a, b, and c. By considering the x4 term in the expansion of the right side, we find a = 1. By
considering the constant term in the expansion on the right side, we have 34c = —136, from which we find c = -4.
Finally, we consider the x3 term on both sides. We have 6ax3 + bx3 on the right, which must equal the 9x3 term on
the left. We already have a - 1, so b = 3. We must check that the other coefficients match up. Checking reveals
that we do indeed have
(x2 + 6x + 34)(x2 + 3x - 4) = x4 + 9x3 + 48x2 + 78x - 136.
Therefore, the other two roots of the quartic are the roots of the quadratic x2 + 3x - 4 = (x - l)(x + 4). So, the roots
of the quartic other than -3 + 5/ are -3 - 5i, 1, and -4
6.47
(a) We are given that the polynomial has five distinct roots, but a polynomial of degree 6 has six roots (counting
repeated roots as different roots), so one of the roots must be repeated.
(b) Since the coefficients of the polynomial are real, the nonreal roots must come in conjugate pairs. The nonreal
roots 2 - i and 2 + z already form a conjugate pair, as do the nonreal roots 1 - i and 1 + i. We also know that
1 is a root, so the sixth and final root must be real. Therefore, the double root is [T].
6.48 We have Y\i = -b + ai and r2i = b + ai. Since the polynomial has real coefficients, the complex conjugate of
r\i = —b + ai, namely —b - ai, must also be a root. Since a and b are nonzero, -b - ai cannot be equal to r2i = b + ai.
Similarly, the conjugate of r2i - b + ai, namely b - ai, must also be a root. We have shown that the polynomial must
have at least four roots. Furthermore, the polynomial
(x + b- ai){x + b + ai)(x - b - ai)(x - b + ai) = [(x + b)2 + a2][(x - b)2 + a2]
has real coefficients, and its roots are -b + ai, -b - ai, b + ai, and b - ai. Therefore, the minimum degree of the
polynomial is [T].
Challenge Problems
6.49 The powers of i repeat every four terms: z° = 1, z1 = z, z2 - -l, i3 = —z, z4 - 1, z5 = z, z6 = -1, z'7 = — i, z8 = 1,
and so on. Hence, the given sum becomes
i + 2 z'2 + 3z3 + 4z'4 + • • • + 64z64

— z — 2 — 3z + 4 + 5z — 6 — 7z + 8 + 9z — 10 — llz + 12 + • • ■ + 61z — 62 — 63z + 64
= (-2 + 4- 6 + 8-10 + 12 +-62 + 64) + (1 — 3 + 5 — 7 + 9-11-f-+ 61- 63)z
= [(-2 + 4) + (-6 + 8) + (-10 + 12) + • • • + (-62 + 64)] + [(1 - 3) + (5 - 7) + (9 - 11) + • • • + (61 - 63)]/
= 2 • 16 - 2 • 16/ 32 - 32/
6.50
(a) Taking the conjugate of the given equation, we get bz + az = c. We can view the equations
az + bz = c,
bz + az = c
116
Challenge Problems
as a system of equations in z and z, and solve for z accordingly. Multiplying the first equation by a and the
second equation by b, we get
aaz + abz = ac,

bbz + abz = be.
Subtracting the second equation from the first gives us
(aa - bb)z - ac - be.
We have aa - bb = \a\2 - \b\2, and we are given that |a| t \b\. Hence, we may divide both sides by aa - bb, to
get
ac - be
z = —-=.
aa - bb
We multiplied by a and by b, and one of these might be 0. So, we have to check our solution and make sure
. . , T, ac-bc . ac- be
it isn t extraneous. If z = —-=, then z = —-=, so
aa - bb aa - bb
, _ aac - abc abc - bbc aac - bbc

az + bz - —z-— h—z-=~ = —z-— = c.
aa - bb aa - bb aa - bb
(b) Let z - x + yi, where x and y are real numbers. Then the equation az + bz = c becomes
(-1 + 3z')(x + yi) + (1 + 3i)(x - yi) = i

« -x - yi + 3xi -3y + x - yi + 3xi + 3y - i
<^> -2yz + 6 xi = i.
Solving for y, we find y -3x-\. Thus, the solutions are of the form x + {3x - z for all real numbers x.
This represents a line passing through the point corresponding to - \i with slope 3.
6.51 Let z — a + bi, where a and b are real numbers. From the equation |z - 1| = |z + 3|, we obtain
| (a — 1) + bi\ — | (a + 3) + bi\
=> y/(a - l)2 + b2= y/{a + 3)2 + b2
=> (a - l)2 + b2 - (a + 3)2 + b2
=> a2 - 2a + 1 + b2 - a2 + 6a + 9 + b2
8 a = -8
a = -1.
Then from the equations |z — 11 = |z — z| and a - —1, we obtain
|-2 + bi\ \-l + (b-l)i\

y(-2)2+f>2 yc-D2+(b - if
4 + b2 1 +(b- l)2
4 + b2 1 + b2 - 2b + 1
-2
-1.
Therefore, z = a + bi = -1-z
117
We can interpret the problem geometrically as follows. The equations |z - 1| = |z + 3| = |z - i\ state that the
complex number z is equidistant from the numbers 1, -3, and i. Hence, z is the circumcenter of the triangle
formed by the complex numbers 1, -3, and z in the complex plane. From |z - 1| = |z + 3|, we find that z is on the
perpendicular bisector of the segment connecting 1 and -3 on the complex plane. This perpendicular bisector
is Re(z) = -1, so the real part of z is -1. From |z — 1| = |z — z'|, we find that z is on the perpendicular bisector of
the segment connecting 1 and z on the complex plane. This line is simply Re(z) = Im(z). Combining this with
Re(z) = -1, we have z = -1 - i, as before.
6.52 The first inequality is satisfied when z is inside or on the circle centered at -4 + 4i with radius 4 x[2, and the
second inequality is satisfied when z is outside or on the circle centered at 4 + 4i with radius 4 V2. Hence, the region
of interest is the set of points that lie inside or on the first circle but outside or on the second circle, as shown below.
Let A be the point -4 + 4i, the center of the first circle, and let B be the
point 4 + 4i, the center of the second circle. Then
AB = |(4 + 4z) - (-4 + 4z')| = |8| = 8.
Let M be the midpoint of AB, so AM = BM = AB 12 = 4.
Let the two circles intersect at P and Q. Since PQ is a common chord

of the two circles, PQ is perpendicular to AB. Furthermore, M is also the
midpoint of PQ.
By the Pythagorean Theorem, AM2 + QM2 = AQ2, so 16 + QM2 = (4 a/2)2 = 32. Therefore, QM2 = 16, so
QM = 4. Then PQ = 2QM = 8.
Now, to find the area of the shaded region, we can take the area of the first circle and subtract two times the
area of the circular segment cut off by PQ. To find the area of the circular segment cut off by PQ, we can find the
area of sector PAQ, and subtract the area of triangle PAQ.
Since AM = QM and /AMQ = 90°, we have /-QAM - 45°. Also, AM - PM and ZAMP = 90°, so /PAM - 45°.
Therefore, /PAQ = 45° + 45° = 90°. Hence, the area of sector PAQ is one quarter the area of the circle, or
|7t(4 y/l)2 = 8tz. The area of triangle PAQ is \AJM ■ PQ - \ • 4 • 8 = 16. Hence, the area of the circular segment cut
off by PQ is 8n - 16.
Finally, the area of the shaded region is 7i(4 V2)2 - 2(87i - 16) = 1671 + 32
6.53 To deal with the appearances of i in the coefficients, we let y = ix. Then the given cubic becomes
zy3 - 8y2 - 22iy + 21 = z'(zx)3 - 8(ix)2 - 22i(ix) + 21= z'(-z'x3) - 8(-x2) - 22i(ix) + 21 = x3 + 8x2 + 22* + 21.
We look for roots of this new polynomial. First, we look for integer roots. If there are any integer roots, then
they must be negative (because all the coefficients are positive) and among the factors of 21 (due to the Rational
Root Theorem). Checking these values, we find that the cubic becomes 0 for * = -3. So by the Factor Theorem,
* + 3 is a factor of the cubic. Dividing by x + 3, we find
x3 + 8x2 + 22x + 21 = (x + 3)(x2 + 5x + 7).
The roots of the quadratic are ■5±^3. Since y = ix, the solutions of the original cubic are
y--
6.54 Let w, x, y, and z be the vertices of a parallelogram in the complex plane, where w and y are opposite
vertices, so x and z are also opposite vertices. In every parallelogram, the diagonals bisect each other; in other
118
Challenge Problems
words, the midpoints of each diagonal coincide. Therefore,
w+y x+z
2 2 y
Thus, the sum of w and y is equal to the sum of x and z.
Conversely, let the sum of two of w, x, y, and z be equal to the sum of the other two. Suppose w + y = x + z.
Then (w + y)/2 = (x + z)/2. This means the midpoint of the line segment between w and y and the midpoint of the
line segment between x and z coincide. In other words, in the quadrilateral formed by the complex numbers w, x,
y, and z, the diagonals bisect each other. Hence, the quadrilateral formed by these four points is a parallelogram.
Make sure you see why we need to include both parts of this proof.
6.55 Let z = To show that z is real, it suffices to show that z = z (as described in the text), so we compute
the conjugate of z. Since \u\ = 1, we have uu = \u\2 = 1, so u = 1/u. Similarly, v = l/v. Then
__ ( 4 uv \ _ 4 uv _ 4 u-v _ 4• \ • 4
V(u + i?)2/ (u + z;)2 {u + v)2 fi + i\2'
\u V)
Multiplying the numerator and denominator by u2v2, this becomes z = = z. Therefore, z = 4uv/(u + v)2 is
real.
Now, let w = U-~-V. To show that w is imaginary, it suffices to show that w = —w. We have
u-v ■
- _ (U + V\ = U+v u+v l + l = l + l ' uv = v + u =

\ u — V/ U-v U — V i-i 1_1 UV v—u
U V U V
Therefore, w - (u + v)/(u - v) is imaginary.
6.56 We have seen the expression x2 + y2 arise when computing the reciprocal of the complex number z = x + yz,
where x and y are real numbers:
1 1 x - yz _ x - yz
z x + yz (x + yz)(x - yz) x2 + y2
Thus, it may be possible to express the given equations in terms of the complex number z = x + yz. We are given
that x/(x2 + y2) = 33x - 56y and 56x + 33y = -y/(x2 + y2), so
x y -i = (33x - 56y) + (56x + 33y)z.

x + yz x2 + -iyz
yZ yZ -vZ _L i /Z
x2 + y2
But (x + yz)(33 + 56z) = 33x + 56xz + 33yz - 56y = (33x - 56y) + (56x + 33y)z, so using our expression for from
above, we have
1 - = (33x - 56y) + (56x + 33y)z = (x + yz)(33 + 56z).
x + yi
Hence,
33 - 56z 33 - 56z 33 - 56z
(x + yz')2 =
33 + 56z _ (33 + 56z)(33 - 56z) _ 4225 652 '
Therefore, [65(x + yz')]2 = (65x + 65yz')2 = 33 - 56z.
Let a = 65x and b = 65y, so (a + hi)2 = 33 - 56z. Expanding, we get fl2 - b2 + 2abi = 33 - 56z. Equating real parts
and equating imaginary parts, we obtain the system of equations
a2 -b2 - 33,
2ab - -56.
119
From the second equation, we have b = -28/a. Substituting into the first equation, we get a2 - 7-jr - 33.
Rearranging, we get aA - 33a2 — 784 = 0, which factors as (a2 - 49)(a2 + 16) = 0, so (a — 7){a + 7)(a2 + 16) = 0. Since a
is real, we have a = 7 or a = -7.
If a = 7, then b = -28/7 = -4, x = a/65 = 7/65, and y = b/65 = -4/65, so \x\ + |y| = 11/65. If a = -7, then
b = -28/(-7) = 4, x = a/65 - -7/65, and y = b/65 = 4/65, so |x| + |y| = 11/65. In either case, |x| + |y| = 11/65
6.57 We compute the first few terms of the sequence:
Z2 = z\ + i = i,
Z3 = z\ + i = -1 + z,
Z4 = z2 + i = (-1 + i)2 + i = 1 - 2z - 1 + z = —z,
Z5 = z| + i = (-z)2 + z = -1 + z,
Z6 = z2 + z = (-1 + z)2 + z = 1 - 2z -1 + z = -z,
and so on.
We observe that the sequence repeats every two terms beginning with Z3. We know that this pattern persists,
because each term in the sequence depends only on the previous term. We conclude that Zui = Z3 = —1 + z.
Therefore, the distance between zui and the origin is |-1 + z'| = yq-1)2 f l2 = V2
6.58 We can produce w\ + w\ + w2 by squaring the first given equation, which gives
w\ + w\ + w\ + 2(W\W2 + W2W3 + W\W3) = 0.
Solving for iv\ + w\ + w\ gives

w\ + w\ + W2 = -2{W\ZV2 + W2W3 + W1W3).
Now, we wish to show that W\W2 + W2W3 + IV1W3 - 0. We let t ~ W1W2 + W2W3 + W1W3. We know we'll need to
use the facts that |wi| = |z^2| = IW3I = 1, which we can write as W\W{ = W2W2 - W3W3 = l2 = 1, to get rid of the
magnitudes. To make use of these equations, we consider tt:
ft = {W\W2 + W2W3 + WiW3)(W\lV2 + W2W3 + W1W3) - (XV2W2 + IV2W3 + lV-ilV3)(lUiW2 + W2W3 + IV1IV3)
= (WiW\)(W2U>2) + W$W21V2)ZV3 + (WiUf)7V2W3
+ (W2TV2)uflV3 + (W2ZV2)(1V3ZV^) + W/lV2(W3llf)
+ (li>iWi)W2lV3 + WiW2(W3W3) + (WiW$(W3ZV3).
We have W\W\ = W2W2 = W3W3 = 1, so
tt = 3 + W1W3 + ZV2W3 + W\W3 + W\W2 + ZV2W3 + wfwi
- 3 + W3(Wi + W2) + Wi{lV2 + TV3) + wifvi + 2^3).
Since u>\ + u>2 + W3 = 0, we can write the three expressions in parentheses as -w3, -W\, and -io2, respectively, and
we have
tt - 3 - W3W3 - WiW\ - W2W2 = 3-1 — 1 — 1 = 0.
Therefore, we have |f| = 0, so t = 0, and we conclude that w\ + w\ + iv2 = 0.
6.59 From the equation \a + bi\ = 8, we get Vzz2 + b2 = 8, so a2 + b2 = 64.
We are also given that the image of each point in the complex plane is equidistant from that point and the
origin. In other words, for all complex numbers z, we have |/(z) - z| = |/(z)|. Substituting /(z) = (a + bi)z, we get
\{a - 1 + bi)z\ - |(a + bi)z\, which we can rewrite as |zz - 1 + bi\ ■ \z\ = |a + bi\ ■ |z|.
120
Challenge Problems
Since this relation holds for all complex numbers 2, the coefficients of |z| must be equal, so
|a — 1 + bi\ = | a + bi\ => \J (a — l)2 + b2 = Va2 + b2 (a - l)2 + b2 = a2 + b2 => a = -.

2
Hence, b2 = 64 - a2 = 64 - Q)2 255

4
Can you find a geometric approach to this problem?
6.60
1 _ 1
(a) If 2 =£ 0, then f(f(z)) = /
TTl ~ 1/2
(b) The problem asks us to find what curve f(z) traces, so we let w = f(z) and try to find an equation for w
that doesn't include 2. Since w - f(z) = 1/z, we have z — 1/w. Taking the conjugate of both sides, we get
2 = 1/w, so 2 = 1/w. Substituting these expressions into the equation (1 + 2i)z - (1 - 2i)z = /, we get
1 + 2/ _ 1-2/
(1 + 2i)w - (1 - 2i)w = iww.
w zv
Let w = a + hi, where a and b are real numbers. Then the equation above becomes
(1 + 2i)(a +
hi) - (1 - 2i)(a - bi) = i(a + bi)(a - bi)
=> (a- 2b) + (2 a + b)i - (a- 2b) + (2a + b)i - (a2 + b2)i
=> (4 a + 2 b)i = (a2 + b2)i
=> la + 2b = a2 + b2
=> a2 - la + b2 - 2b -0
=> a2 - la +1 + b2 - 2b + 1 = 5
=> (a- 2)2 + (b- l)2 = 5.
Hence, it appears that w traces a circle centered at 2 + / with radius V5. However, we must be careful to
note that w cannot be 0, since there is no complex number 2 for which 0 = 1/z. So, w traces the entire circle
centered at 2 + / with radius V5, except the origin.
6.61 We have seen the expression a2 + b2 arise when computing the reciprocal of the complex number 2 = a + bi,
where a and b are real numbers:
1 _ 1 _ a - bi _ a-bi
2 a + bi (a + bi)(a - bi) a2 + b2'
Thus, it may be possible to express the given equations in terms of the complex number z = a + bi.
We see that the first equation contains the term a, which is the real part of a + bi. The second equation contains
the term b, which is the imaginary part of a + bi. Furthermore, the first equation contains the term jzfp, which is
the real part of 1/z, as seen above, and the second equation contains the term which is the imaginary part
of 1/2.
Hence, by taking the expression in the first equation as the real part of a complex number, and the expression
in the second equation as the imaginary part, we get
a + 8b \ 8 a-b\ a - bi 8 (b + ai) 8 (b
+ ai)
2 + 0/ = + i = a + bi + (♦)
a2 + b2) a2 + b2) a2 + b2 + a2 + b2 a2 + b2 '
8(b + ai) .
where 2 = a + bi. This leaves us with the task of expressing z—— m terms of 2.
a2 + b2
, .s . ,1 a - bi 8(b + ai) 8i(a - bi) 8/ . 1 8/
Since i(a-bi) = b+ai and - = -=—t=, we have —=— —-—- = —. Thus, we can write (A) as 2H-1— = 2.
2 a2 + b2 a2 + bz a2 + b2 2 22
121
Multiplying both sides by z, we get z2 + 1 + 8i - 2z. Rearranging this equation gives z2 - 2z + 1 = -8i, so
(z - l)2 = -Si. Hence, we must find the numbers whose squares are -8i.
Let w - c + di, where c and d are real numbers and iv2 = -Si. Then, we have w2 = (c + di)2, so c2-d2 + 2c. di = -Si.
Equating the real parts and equating the imaginary parts, we get the system of equations c2-d2 - 0 and led = -8.
From the second equation, we have d - -4/c. Substituting this expression for d into the first equation, we get
16
c2 = c4 = 16 c = ±2.
(We cannot have c = 2i or c = -2i because c is real.) The corresponding values of d = -4/c are t2. (The + symbol
indicates that d - -2 goes with c - 2 and d = 2 goes with c = -2.) Therefore, the square roots of -Si are 2 - 2i and
-2 + 2 i.
Hence, z - 1 = 2 - 2/ or z - 1 = -2 + 2i, so z = 3 - 2i or z = -1+2i. The corresponding values of (a, b) are (3, -2)
and (-1,2). We check that both of these solutions work, so the solutions are (a, b) = (3, -2) and (-1,2)
122
CHAPTER 7 l
l Trigonometry and Complex Numbers
7.1.1
(a) The real part is 0 and the imaginary part is -32, so r = \J02 + (-32)2 = 32. We seek an angle 0 such
that 32 cos 0 = 0 and 32 sin 0 = -32, which means cos0 = 0 and sin0 = — 1. We can take 0 = — , so
Ttt * •
-32z = 32 ( cos y + z sm ¥)
(b) The real part is 2 and the imaginary part is -2, so r = yjl2 + (-2)2 = V8 = 2 V2. We seek an angle 0 such
that 2 V2 cos 0 = 2 and 2 V2 sin 0 = -2, which means cos 0 = ~ and sin 0 = —4^. We can take 0 = ^, so
2-2 z = 2 V2 ( cos y + z sin ~
(c) The real part is 2 V3 and the imaginary part is 6, so r = \J(2 V3)2 + 62 = V48 = 4 V3. We seek an angle 0
such that 4 V3 cos 0 = 2 V3 and 4 V3 sin 0 = 6, which means cos 6-1 and sin 0 = y- We can take 0 =
so 2 V3 + 6z = 4 V3 (cos f + z sin
(d) The real part is | and the imaginary part is so r = \/(1/4)2 + (- V3/4)2 = VT/4 = |. We seek an angle
0 such that ^ cos 0 = | and | sin 0 = —^, which means cos 0 = \ and sin 0 = —y- We can take 0 = y so
1 _ &i = cos y + * sin x)

4 4 4
7.1.2
(a) 6 cis 150° = 6 cos 150° + 6z sin 150° = 6 • (- y) + 6z • | = -3 V3 + 3z
(b) 8 cis 3780° = 8 cis 180° = 8 cos 180° + 8z sin 180° = 8 • (-1) + 8z • 0 = -8
(c) 2 cis y = 2 cos y + 2z sin y = 2 • (-5) + 2z • y = + V3z
(d) 9 cis y = 9cos y + 9zsin y = 9 • (- y) + 9z • y = ~
7.1.3 We showed in the text that cis x cis y = cis(x + y) for any x and y, so cis cis(n - jS) = cis[/5 + (a - /?)] = cis a,
which means §§f = cis(n - 0). Hence, we have f = ; cis(n - jS).
123
CHAPTER 7. TRIGONOMETRY AND COMPLEX NUMBERS
We also could have tackled this problem as follows:
vo r(cos a + z sin a) _ r(cos a + i sin a) cos (1 - i sin jS

z s(cos jS + i sin jS) s(cos /J + z sin j3) cos j3 - z sin /?
r (cos a cos jS + sin a sin j3) + (sin a cos /? - cos a sin j3)z
s cos2 j5 - z2 sin2
r cos(a - B) + i sin(« - B) r . .
- •-—r-5-— = - cis(a - p)
s cos2 /3 + sin2 jS s
7.1.4
(a) For the first factor, we can visualize a right triangle with legs 6 V3 and 18. Since one leg is V3 times
the other, this triangle is a 30-60-90 triangle. Its hypotenuse is twice the shorter side, so the hypotenuse
is 12 V3. Therefore, the magnitude of 6 V3 + 18/ is 12 V3, and we have 6 a/3 + 18/ = 12 V3 Q + =
12 V3 (cos | + i sin ^.
Similarly, a right triangle with legs 2 V3 and 6 is a 30-60-90 triangle with hypotenuse 4 V3, and we have
-2 V3 + 6z = 4 V3 = 4 a/3 (cos 23- + z sin .
So, we have
(6 V3 + 18z)(—2 V3 + 6z) = (12 V3) ^cos ^ + z sin ?) (4 V3) (cos ^ + z sin Û
/ 7i . . 7i\ / 2n 2n
. . Zn
= 48-3 cos — + z sin — cos — + z sin —
V 3 3/
3 V 3 3
71 271 \ n 2n
= 144 cos I + z sin
3 + T/ 3 + ~3~
-144
All these computations can be done in your head—find the magnitudes using 30-60-90 triangles, and the
arguments using your understanding of cosine and sine. Then, multiply the magnitudes and add the
arguments.
(b) As in part (a), we use our understanding of special right triangles to find the magnitudes, and our under¬
standing of cosine and sine to find the arguments. Considering 45-45-90 right triangles, we see that the
magnitudes of the three factors are 4 V2, 3 V2, and 10 V2, respectively. Each has an argument of ^ for
some odd k. Since the argument of 4 + 4z must be in the first quadrant, it is |. The argument of -3 - 3z
is in the third quadrant, so it is The argument of -10 + 10/ is in the second quadrant, so it is ^p.
Therefore, the magnitude of the product is (4 V2)(3 V2)(10 V2) = 240 V2, and the argument of the product
isp + Î + ^ = Î/ and the product is
( 9n 9n\
240 V2 cos — + z sin — = 240 + 240z
V 4 4 7
7.1.5 Since cis0 = cos0 + zsin0 and cis(-@) = cos(-0) + zsin(-@) = cos0 - zsin0, we see that cis0 = cis(-@)
if and only if sin 6 - - sin 6. But sin 9 - - sin 9 if and only if sin 0 = 0. Finally, sin 0 = 0 if and only if 0 is a
multiple of n .
See if you can also solve this problem geometrically. How are cis 0 and cis(-0) related to the point correspond¬
ing to 1 in the complex plane?
124
Section 7.2
7.1.6 For 0 = 30°, we have cos 50 = cos 150° = -4p, and
16 cos5 0-20 cos3 0 + 5 cos 0 = 16 |

V3
-20
:*) 2
32 V3 3 V3 5V3 9 V3 15 V3 5 V3 V3
- 16- -20 —-— +
32 8 + 2 2 ~Y~ 2 '
Thus, the identity holds for 0 = 30°.
7.1.7 By de Moivre's Theorem, cos 70 + / sin 70 = (cos 0 + /sin 0)7. By the Binomial Theorem,
(cos 0 + / sin 0)7 = cos7 0 + 7/ cos6 0 sin 0 - 21 cos5 0 sin2 0 - 35/ cos4 0 sin3 0
+ 35 cos3 0 sin4 0 + 21/ cos2 0 sin5 0-7 cos 0 sin6 0 - / sin7 0.
We know that this equals cos 70 + / sin 70, so equating the real part of this to the real part of the expansion above,
we find
cos 70 = cos7 0-21 cos5 0 sin2 0 + 35 cos3 0 sin4 0-7 cos 0 sin6 0.
We want to express this only in terms of cos 0, not sin 0. Substituting sin2 0 = 1- cos2 0, we get
cos7 0-21 cos'"’ 0 sin2 0 + 35 cos3 0 sin4 0-7 cos 0 sin6 0
= cos7 0-21 cos5 0(1 - cos2 0) + 35 cos3 0(1 - cos2 0)2 - 7 cos 0(1 - cos2 0)3
= cos7 0-21 cos5 0 + 21 cos7 0 + 35 cos3 0(1 — 2 cos2 0 + cos4 0)
- 7 cos 0(1 - 3 cos2 0 + 3 cos4 0 - cos6 0)
= 64 cos7 0 - 112 cos5 0 + 56 cos3 0-7 cos 0 .
7.1.8 Since cos(90° - 0) = sin 0 and sin(90° - 0) = cos 0, we have
sin t + / cos t = cos(90° - t) + i sin(90° — f) = cis(90° - /),
so by de Moivre's Theorem,
(sin t + / cos t)n = cis''(90° - f) = cis[n(90° - /)] = cis(90°n - nt).
Likewise, sin nt + / cos nt = cis(90° - nt). Thus, the given equation becomes cis(90°n - nt) = cis(90° - nt).
This equation holds for all t if and only if two the arguments differ by a multiple of 360°. In other words, if
and only if 90°n - 90° = 360°k for some integer k. Hence, n - 1 = 4k, so n = 4/c + 1. The positive integers less than
or equal to 1000 that are of this form are 1, 5, 9,..., 997, of which there are 250
7.2.1
(a) The real part is 0 and the imaginary part is -32, so r - yj02 + (-32)2 = 32. We seek an angle 0 such
that 32 cos 0 = 0 and 32 sin 0 = -32, which means cos 0 = 0 and sin0 = -1. We can take 0 = so
-32/ = 32e3ni/2
125
(b) The real part is -8 and the imaginary part is 8 V3, so r - \](-8)2 + (8 V3)2 = V256 = 16. We seek an angle
0 such that 16 cos 0 = -8 and 16 sin 0 = 8 V3, which means cos 0 = - \ and sin 0 = . We can take 0 =
so -8 + 8 V3z = 16e2ra/3
(c) The real part is — ^ and the imaginary part is-^p, sor=y(--^) + (“^) = \[\ ~ 3- We see^ ^ anSê
0 such that | cos 0 = -^ and ^ sin 0 = which means cos 0 = and sin 0 = 4^. We can take 0 =
1 .3ra'/4
so + %= 3e
(d) The real part is 2 Vl5 and the imaginary part is 2 V5, so r = \/(2 Vl5)2 + (2 V5)2 = V80 = 4 a/5. We seek an
angle 0 such that 4 a/5 cos 0 = 2 a/15 and 4 V5 sin 0 = 2 V5, which means cos 0 = ^ and sin 0 = |. We can
take 0 = f, so 2 Vl5 + 2 V5z = 4 V5em/6
7.2.2
(a) e4ra = cos 47i + z sin 471 = [T].
(b) e57n/4 = cos ^jr + i sin = _ VI

2
_ YI}-
2 4
(c) e 5ttV3 = e 5m/3+2m = era/3 = cos 5 + z sin 5 = 21 T

+ VI*
2
(d) e17™/3 = e17ra/3 4ra = e5™/3 = cos ~ + z sin ^ = YT

2 2 Z
7.2.3 We have = cos 0 + z sin 0 and e 16 = cos(-0) + z sin(-0) = cos 0 - z sin 0. Subtracting the second equation
from the first equation, we get etd - e~l° - 2/ sin 0, so sin 0 = e‘e-£>e.
7.2.4 We have
etx _ cosx + zsinx _ (cosx + zsinx)(cosy - zsiny)

e'y cos y + i sin y (cos y + z sin y)(cos y - z sin y)
cos x cos y + sin x sin y + (sin x cos y - cos x sin y)z
= cos x cos y + sin x sin y + (sin x cos y - cos x sin y)z.
cos2 y + sin2 y
We also have ^ = e1^ yS> = cos(x - y) + z sin(x - y). Equating the real part of this to the real part of our expression
above for gives cos(x - y) = cos x cos y + sin x sin y, and equating the imaginary parts gives sin(x - y) =
sin x cos y - cos x sin y.
7.2.5
(a) By considering 45-45-90 right triangles and our knowledge of sine and cosine, we can see that 8+8/ = 8 y/2ent/4
and -3 + 3/ = 3 y/le3™14, so their quotient is |g(”'/4)-(3n!/4) = =
(b) A right triangle with legs 4 V3 and 12 has one leg that is V3 times the other, so it is a 30-60-90 triangle with
hypotenuse twice the smaller leg, which means the hypotenuse is 2 • 4 V3 = 8 a/3. Therefore, -12 - 4/ V3
has magnitude 8 V3. Its argument is the third quadrant angle whose sine equals ~\, so its argument is
Similarly, 3 + 3/ V3 has magnitude 6 and argument |. So, we have
-12 —4iV3 8V3e™* 4 V3 g„761-(m/3, 4V3 ,„i/6 4V3 /_V3 1A

3 + 3/V3 3 3 3 ( 2 2 )
126
Section 7.3
7.2.6
(a) First we express 1 -z in exponential form. The real part is 1 and the imaginary part is -1, so r = \J l2 + (-1)2 =
V2. We seek an angle 0 such that V2 cos 0 = 1 and V2 sin 0 = -1, which means cos 0 = ^ and sin 0 = — ^.
We can take 0 = ?f, so 1 - i = V2c7tt'/4. Hence, (1 - z')16 = (V2c7ra/4)16 = 28e28m = 256
(b) First we express -2 V3 + 6z in exponential form. The real part is -2 V3 and the imaginary part is 6, so
r = \/(~2 V3)2 + 62 = V48 = 4 V3. We seek an angle 0 such that 4 V3 cos 0 = -2 V3 and 4 V3 sin 0-6,
which means cos 0 = and sin 0 = We can take 0 = so -2 V3 + 6z = 4 V3e2m/3. Hence,
(-2 V3 + 6/)8 = (4 y/3e2ni/3)8 = 48 • 3V6ti</3 = 48 • 34 -2654208 - 2654208* V3
7.2.7 We express -3 + a/3z in exponential form. The real part is -3 and the imaginary part is V3, so r =
y/(~3)2 + (V3)2 = Vl2 = 2 V3. We seek an angle 0 such that 2 V3 cos 0 = -3 and 2 V3 sin 0 = V3, which means
cos 0 = --y1 and sin 0 = We can take 0 = so -3 + a/3z = 2 y/3e5nl/6. Hence,
5n7T . . 5n7i\
(-3 + V3z')n = (2 yf3)ne5nni/6 = (2 V3)M cos —— + z sin -
■
This is a real number if and only if sin ^ = 0. But sin = 0 if and only if ~ is a multiple of tc, i.e. if =
for some integer k, which means k = 5n/6. Since k is an integer, n must be divisible by 6.
Conversely, if n is a multiple of 6, say n - 6m, then
(-3 + V3z)n = (2 V3e5Tn'/6)6,n = (2 y[3)6me5mni = (2 V3)6m(-l)5m,
which is a real number. Therefore, (-3 + V3z)n is a real number if and only if n is a multiple of 6
7.2.8 Since dividing one exponential expression by another is easy, we try to show that
em + elP (a -
2e!'(a4f)/2 " cos ^“2“ J •
Multiplying both sides of this by 2e‘â+^/2 will give the desired equation. (This expression cannot be 0 because
e,e t 0 for all angles 0.)
We see that
1 / gia gift \ gi(a-jS)/2 + gf(/J-a)/2 i(a~P)/2 + g-Ha-l3)/2
eia + e'P a - ]8
= COS
2ei(a+P)/2 2 + gka+^)/2 J ~ 2
so = 2 cos e'<'°‘+®12-
7.3.1 The solutions of x8 = 1 are the eighth roots of unity, namely
127
The primitive eighth roots of unity are the eighth roots of unity of the form e2km/8 for which gcd(/c, 8) =
1. Since 8 = 23, we have gcd(fc, 8) = 1 if and only if k is odd, so the primitive eighth roots of unity are
g2ra'/8 e6ni/8 g10ni/8 an(j gl4m'/8
7.3.2 Writing -64 in exponential form, we get -64 = 64en!. Therefore, the roots of the equation x6 = -64 = 26enl
are 2eni/e, 2e3m/6, 2e5m/6,2e7ni/6, 2e9ni/6, and 2eUni/6.
The real part of a complex number is positive if and only if its argument is in the first or fourth quadrant (or
if the argument is a multiple of 2n). The only roots of x6 = -64 that have such an argument are 2en^6 and 2enm/6,
and their product is 4em/6+llra/6 = 4e27n = [4].
7.3.3 The sixth roots of unity are e° = 1, g2™/6 = gra/3,e4m/6 _ e2m/3,e6m/6 _ gi _ _4^g8ra/6 _ g4m/3/ and g10m/6 = e5ra/3.
We compute co + co2 for each such root:
1 + V = 2,
ni/3 27n/3 71 . . n 2n . . 2n 1 . V3 1 . V3 rz.
e 1 + e ' = cos — + z sm — + cos — + z sin — = - + 1 • —— - +1 • —— - V3z,
3 3 3 32222
2m73 ini 13 2n . . 2n 4tI . . 4 71 1 . V3 1 . V3
e ' + e 1 = cos —- + z sm —- + cos — + zsm— = --+/• —— - - i • - = -1,
3 3 3 32222
-1 + (-1)2 = 0,
4ra/3 8tii73 471 .. 471 871 .. 871 1 . V3 1 . V3
e 1 +e 1 - cos — + z sin — + cos — + z sm — = -- - z • —-- + z • —— = -1,
3 3 3 32222
5711/3 10711/3 571 ■ • 571 1071 . . 1071 1 . V3 1 . V3 yr.
r 1 +e 1 - cos — + z sm — + cos —— + z sm —— = - — z • —— - - z • - - V3z.
3 3 3 32222
Thus, the possible values of co + co2 are 2, 0, -1, V3z, and - V3z
7.3.4 We tackle both parts at once.
Solution 1: Consider each of the \2lh roots of unity. The 12th roots of unity are the values of e2™^12 for k =
0,1,2,..., 11. As explained in the text, if gcd(/c, 12) = 1, then e2ntk/u is a primitive 12th root of unity. That covers
k = 1,5,7,11, so there are 4 primitive 12th roots of unity.
If k = 2, we have e2mk/u = e271^6, which is a primitive 6th root of unity, since (e2m/6)6 = 1 but the argument
of (e2™/6)" is between 0 and 2n for 1 < n < 5. The same holds for k = 10, which we can see by noting that
£,27n-io/i2 _ ^2711-12/12^711-2/12 _ |^711/6_ ^2tt!-io/12^6 _ ^ an£j no smaqer positive integer power of e2™-10/12 equals
1. So, there are 2 primitive 6th roots of unity among the 12th roots of unity.
Similarly, we find that when k - 3 or k = 9, then we have (e271^/12)4 = but no smaller positive power of e2nlkl12
equals 1, so there are 2 primitive 4th roots of unity among the 12th roots of unity.
We also find that if k = 4 or k = 8, then (e2™^12)3 = 1, but no smaller positive power of e2mk/12 equals 1, so there
are 2 primitive 3rd roots of unity among the 12th roots of unity.
Finally, if k - 6, we have (g2Tnfc/12)2 = \ but (e2mk^2)1 ± 1, and if k = 0, then (e271^/12)1 - So there is 1 primitive
first root of unity and 1 primitive square root of unity among the 12th roots of unity.
Solution 2: Use a little number theory. Let a; be a 12th root of unity. Since co12 = 1, there is at least one positive
integer n such that con = 1. Let k be the smallest such positive integer. Then (ok = 1 and +r = 1. But a)kcf+r = (cok)^ • cor = U ■ cor = cor, so a/ = 1. But r < k - 1, so r cannot be
positive. (By definition, k is the smallest positive integer such that cok = 1.) Therefore, r = 0, which means 12 - kq.
128
Section 7.3
Hence, k must be a divisor of 12.
Conversely, if co is a primitive ftrth root of unity, where ft: is a divisor of 12, then clearly co is a 12th root of unity.
Therefore, the possible values of k are 1, 2, 3,4, 6, and 12
The number of primitive ft:th roots of unity is the number of positive integers less than or equal to k that have
no positive factors in common with k besides 1. Thus, we only have to compute this for each quantity for each
divisor of 12. Doing so, we find that among the 12th roots of unity, we have [T] primitive first root of unity, [T]
primitive square root of unity, [2] primitive cube roots of unity, [2] primitive fourth roots of unity, [2] primitive
sixth roots of unity, and [I] primitive 12th roots of unity. (Note that these account for alll + l+ 2 + 2 + 2 + 4 = 12
12th roots of unity.)
7.3.5 Let co be a primitive (2m)th root of unity, so co2m = 1 and cok t 1 for k = 1,2,..., 2m -1. Then (co2)m = co2m = 1,
so co2 is an mth root of unity. Furthermore, (co2)k - co2k ^ 1 for k = 1,2,..., m — 1, because of our observation above.
Therefore, co2 is a primitive mth root of unity.
7.3.6
(a) Let co - afi. Then co15 — (a-/?)15 = a15j315 = (a3)5(jS5)3 = 1, so co is a 15th root of unity.
(b) Let co = afi. Let k be an integer such that cok = 1. Then (ap)k = akfik = 1. Raising both sides to the power of
3, we get a3kp3k - 1. Since a is a cube root of unity, we have a3k = 1, so fi3k - 1. But /3 is a primitive fifth root
of unity, so 3k must be divisible by 5. Since gcd(3,5) = 1, we know that k must be divisible by 5.
Similarly, from the equation akfik = 1, we get a5kfi5k = 1. Since /3 is a fifth root of unity, it follows that
a5k = 1. But a is a primitive cube root of unity, so 5k must be divisible by 3. Since gcd(3,5) = 1, we know
that k must be divisible by 3.
We have established that k must be divisible by both 3 and 5. Since gcd(3,5) = 1, we know that k must
be divisible by 3 • 5 = 15. Since this holds for any k such that cok = 1, we conclude that k must be a multiple
of 15. Since co15 = 1 and 15 is the smallest multiple of 15, we know that co is a primitive 15th root of unity.
(We can show similarly that if a is a primitive nth root of unity, and /I is a primitive mth root of unity, and
gcd(m,n) = 1, then af3 is a primitive (mn)th root of unity.)
(c) We see that e2m/4: = eml2 is a primitive fourth root of unity, and that e2m,e = eml3 is a primitive sixth root of
unity. So, letting a = eml2 and = em/3, we have
ap _ eni/2 . eni/3 _ gni/2+ni/3 _ g5m'/6.
Since (n/i)12 = (e571^6)12 = eWm = 1, the product afl is not a primitive 24th root of unity.
7.3.7 We need a polynomial whose roots are the twelfth roots of unity. So, we start with the polynomial whose
roots are all of the 12th roots of unity, namely x12 -1 = 0. As a difference of squares, we have x12 -1 = (x6 - l)(x6 +1).
We can factor x6 - 1 as a difference of squares: x6 - 1 = (x3 - l)(x3 + 1). As a difference of cubes, we have
x3 - 1 = (x - l)(x2 + x + 1), and as a sum of cubes, we have x3 + 1 = (x + l)(x2 - x + 1).
We can also factor x6 + 1 as a sum of cubes, x6 + 1 = (x2 + l)(x4 - x2 + 1). Thus, we have the factorization
x12 - 1 = (x - lXx2 + x + l)(x + l)(x2 - x + l)(x2 + l)(x4 - x2 + 1).
The only quartic factor that could possibly fit the problem is therefore x4 - x2 + 1
Proving that this quartic is irreducible is a good deal harder than finding it. Suppose that x4 - x2 + 1 can be
factored over the rational numbers. First, we identify the roots of this polynomial. To do so, we must identify the
roots of the other polynomials in the factorization above.
From our work above, we see that x4 - x2 + 1 is a factor of x6 + 1. Since x12 - 1 = (x6 - l)(x6 + 1), the roots of
x6 + 1 = 0 are the 12th roots of unity that are not 6th roots of unity.
129
The 12th roots of unity are e°, e2m^2, e4nl^2,..., e22n!/12, and the 6th roots of unity are e°, e2ml6r e4m/6,..., eWm/6.
Note that e° = e°, e4mZ12 = e2™/6, e8™/12 = e4n and so on. Thus, the 12th roots of unity that are not 6th roots of
unity are e2ni^2, C6m/12f glOni/12^ eUni/12, e\8m/12' ^22ni/12^
Furthermore, x6 + 1 = {x2 + l)(x4 - x2 + 1). The roots of x2 + 1 = 0 are i = em/2 - e6m/u and -i = e3ra/2 = e18m,n.
Therefore, the roots of x4 - x2 + 1 = 0 are e2m^2, eWmZ12, e14™/12, and e22m/12.
If x4 - x2 +1 factors, then it either factors (at least) as the product of a linear polynomial and a cubic polynomial,
or as the product of two quadratic polynomials. But each root is nonreal, so there cannot be any linear factors
with rational coefficients. Therefore, it must factor as the product of two quadratic polynomials.
If a quadratic polynomial has real coefficients, and it has nonreal roots, then those roots must be conjugate
pairs. Therefore, the roots of one of the quadratic factors must be e2ni/u and e22nt/u. But
e2™112 = eni/6 = cos ^ + *'sin ^

6 6 2 2
^22*1/12 = eUm/6 = cos 112 + isin UH V5

:*/
2
which makes the quadratic
V3 l.\ V3 1\
V3. V3x + 1.
2 2J 2 + 2 y 4 4
Even if we multiplied this quadratic by a nonzero rational number, we couldn't make all the coefficients rational.
Therefore, the polynomial x4 — x2 + 1 is irreducible over the rational numbers.
7.4.1
(a) Since (z + l)(z2 - z + 1) = z3 + 1, the roots of z2 - z + 1 = 0 are the roots of z3 = -1 other than -1. Since
-1 = em is one root of z3 = -1, the other roots are em+2n^3 = e5ml3 and eml3+2'2m^ = e7m/3 = em/3 .
(b) Since (z5 - l)(z5 + 1) = z10 - 1, the roots of z5 + 1 are the roots of z10 - 1 that are not roots of z5 - 1. The roots
of z10 - 1 are e°, e2m/w, e4ml10,..., g18™/10. The roots of z5 - 1 are e°, e2m/5, e4m^5, e6m/5, and e8m/5.
Note that e° = e°, e2nl/5 - e4n7//1°, e4nl^ = e8m/w, e6m/5 = cUn^w, and e8m^ = e16ra/10. Therefore, the roots of
e3ni/5 e7ni/5
z5 + 1 are e2ni/w = eni/5 / cp6ni/l0 __ gl07I//10 _
eni / cpUni/lO _
, and e18nllw = 097n/5
(c) The polynomial z5 + 1 factors as (z + l)(z4 - z3 + z2 - z + 1). The only root of z5 + 1 that is real is -1, so the
degree 4 polynomial whose roots are the nonreal roots of z5 + 1 is z4 - z3 + z2 - z + 1
(d) First, the polynomial x7 +1 factors as (x+1 )(x6 - x5+x4 - x3+x2 - x +1), so the roots of xb - x5 + x4 - x3 + x2 - x+1
are the roots of x7 + 1 other than -1.
From here, we can either find the roots of x7 + 1 using the the same technique as in part (a), or we can be
a little clever. We'll show the clever approach here. Since (x7 - l)(x7 + 1) = x14 - 1, the roots of x7 + 1 are the
roots of x14 -1 that are not also roots of x7 - 1. The roots of z14 - 1 are e°, e2m/u, e4™/14,..., e26ni/u. The roots
of z7 — 1 are e°, e2n^7, e4n^7,..., el2n^7. Note that e° = e°, e2ni^7 = g4™/14, e4niT — g87I!/14/... f e12ni/7 = e2^m/u^
Therefore, the roots of z7 + 1 are e2nû - eni^7, e6mZ14 = e3m^7, e1071'/14 = e5n^7,..., e2677'/14 = e13m/7.
Finally, -1 = em = e7nl17, so the roots of x6 - x5 + x4 - x3 + x2 - x + 1 =0 are
eni/7, e?>nil7f e5ni/7f e9ni/7 gHni/7 an(j el3ni/7
130
Section 7.4
7.4.2 The roots of x - 1 - 0 are the 13th roots of unity, namely 1, co, co2, ..., co12. Also, the polynomial x13 - 1
factors as x13 - 1 = (x - l)(x12 + x11 + • • • + x + 1). Hence, the roots of x12 + x11 +-1- x + 1 = 0 are co, co2,..., co12,
which means
(x - co)(x - co2) • • • (x - co12) = X12 + x11 +-h X + 1.
Taking x = 1, we find (1 - co)(l - co2) ■ ■ ■ (1 - co12) = 13
7.4.3 Putting the expression over a common denominator, we get
20 , ,40 , -60 , ,80 , ,100

no ,30 z60 + z40 + z80 + z20 + z100 + 1 1 + z/u + Z4U + ZbU + Z8U + z
Ziu + -T7T + + — + z50 + =
,10 ■30 ,50 ,50 ,50
From the formula for a geometric series, the numerator is equal to
,120 _ i
20 ,40 , ,60 , ,80 , ,100 _ £_£
1 + ZZU + Z4U + zbU + Z8U + ziUU =
Z20 - 1 '
,
Since z7 = 1 we have z120 = z17'7+1 = (z7)17 • z = z, z20 = z27+6 = (z7)2 • z6 = z6, and z50 = z7'7+1 = (z7)7 • z = z, so
,120
1 + z20 + z40 + z60 + z80 + z100 Z- 1 1 Z— 1 Z— 1
^50 z50(z2o - 1) z(z6 - 1) z7 - z 1 -z —(z - 1) =ED
We also could have made reductions in the powers of z before summing the geometric series. We have
z20 = (z7)2z6 = z6. Treating the other powers of z similarly gives
1 + z20 + z40 + z60 + z80 + z100 _ 1 + z6 + z5 + z4 + z3 + z2

z50 z
Because z is a seventh root of unity besides 1, we have z6 + z5 + z4 + z3 + z2 + z +1 = 0, so z6 + z5 + z4 + z3 + z2 +1 = -z,

which means the fraction above equals -z/z, which is -1.
7.4.4 Let k = | = | = ~a. Then P - | • h- • ^ = 1, so k is a cube root of unity, i.e. k = 1, e2m/3, or e4m/3.
We can write b and c in terms of a by noting that c = ka and b = kc, so b = A:2a. We then have (a + b - c)/(a - b + c) =
(a + k2a-ka)/(a-k2a + ka) = (1 +k2-k)/(l-k2 + k). This is easy to compute for one cube root of unity; for k = 1, this
expression equals 1. We could pound it out for each of the other cube roots of unity, but instead we can make a
clever simplification. If k is a cube root of unity besides 1, then we have 1 + k + k2 = 0, so 1 + A: = -k2 and 1 + k2 = -k.
Making these substitutions in the numerator and denominator of the fraction give
1 + k2 - k _ (1 + k2) - k _ -k-k _ 1
1 - k2 + k ~ (1 + k) - k2 ~ -k2 - k2 ~ k'
Since A: is a cube root of unity, we have P - 1, so | = k2. Therefore, the other two possible values are the squares
of the primitive cube roots of unity. But (e2n,/3)2 = eAm/3 and (<?4m/3)2 = e8m/3 = e2m/3, which means that our other
two possible values are the cube roots of unity.
Therefore, the possible values of are 1, and - TpZ
We also could have simplified the solution even more with the observation that a = kb,b = kc, and c = ka, so
a+b-c _ a+b-c _ a+b-c _f_^_^2

a—b+c kb -kc + ka k(a + b - c) k k
131
7.4.5 We compare the polynomials x7 + x4 + x3 + x + 1 and x4 + x3 + x2 + x + 1, because both polynomials are the
sum of five powers of x, and both have several terms in common. In fact, their difference is (x7 + x4 + x3 + x + 1) -
(x4 + x3 + x2 + x + 1) = x7 - x2, which factors as
x7 - x2 = x^x5 - 1) = x2(x - l)(x4 + x3 + x2 + X + 1).
Therefore,
x7 + x4 + x3 + X + 1 = (x4 + x3 + x2 + X + 1) + (x7 - x2)

= (x4 + x3 + x2 + X + 1) + x2(x - l)(x4 + x3 + x2 + X + 1)
= (x4 + x3 + x2 + x + 1) + (x3 - x2)(x4 + x3 + x2 + X + 1)
= (x3 - x2 + l)(x4 + x3 + x2 + X + 1) .
7.4.6 Since 1 + x + x2 +-H x17 = , we can write
x36 - 2x18 + 1 „17

P(x) = (1 + * + *2 + • ■. +117)2 - x17 = 0^x1 ) 2 - x>7 = -x17 =
x2 - 2x + 1
_ x36 - 2x18 + 1 - x19 + 2x18 - x17 _ x36 - x19 - x17 + 1
x2 - 2x + 1 x2 - 2x + 1
The numerator factors as x36 — x19 - x17 + 1 = x19(x17 — 1) - (x17 - 1) = (x19 - l)(x17 - 1), so
x36 - x19 - x17 + 1 _ x19 - 1 x17 - 1

x2-2x + l x- 1 x-1
The roots of (x19 - l)/(x - 1) = 0 are the 19th roots of unity other than 1, and the roots of (x17 - l)/(x - 1) = 0 are
the 17th roots of unity other than 1. Hence, the n;, in some order, are equal to
12 16 1 | 18
17' 17' 17' 19' 19' 19'
The smallest five a, are a.\ = 1/19, = 1/17, n3 = 2/19, oc\ = 2/17, and n3 = 3/19, so
1 1 2 2_ 3 _ 159
al+a2 + a3 + (Xi + a5 - — + — + — +
17 + 19 " 323
Review Problems
7.33 In each part, we are asked to convert a rectangular form x + yi to a polar form r(cos 6 + i sin 0). As described
in the text, this is equivalent to converting the rectangular coordinates (x, y) to the polar coordinates (r, 6).
(a) We have r = yj(-9)2 + 02 = V8l = 9. We seek an angle 6 such that 9 cos 6 = -9 and 9 sin 0 = 0, which
means cos 0 = -1 and sin 0 = 0. We can take 0 = n and express -9 in polar form as 9 cis n
(b) We have r = \j(V6)2 + ( V2)2 = V8 = 2 V2. We seek an angle 0 such that 2 V2 cos 0 = V6 and 2 V2 sin 0 =
a/2, which means cos 0 = -y and sin 0 = |. We can take 0 = | and express V6 + a/2i in polar form as
2 V2 cis ^ .
6
132
Review Problems
(c) We have r - V82 + 82 = Vl28 = 8 V2. We seek an angle 0 such that 8 y/2 cos 0 = 8 and 8 V2 sin 0 = 8, which
means cos 0 = =r and sin 0 = ~ ■ We can take 0 = j and express 8 + 8i in polar form as 8 V2 cis -7
4
(d) We have r - y (-|) + (-|) = = 4^. We seek an angle 0 such that y cos 0 = -| and Y sin0 =
which means cos0 = ~y and sin0 — —We can take 0 = ^ and express — | - |z in polar form as
V2 571
2 C1SX
7.34
(a) We have
(b) We have 1 V2 V2 •
2
(c) We have 4 V3 + 4 i
(d) We have 20 cis 97n = 20 cis 71 = 20 cos 71 + 20i sin n = -20
J0
7.35 We have
(which is never 0), we get el6 = 1. Then cos 0 + i sin 0 = 1, so cos 0 = 1 and sin 0 = 0. Hence, the solutions in 0 are
integer multiples of 2tc
We also can tackle this problem by noting that cos 0 + i sin 0 = cos 20 + i sin 20 if and only if cos 0 = cos 20 and
sin 0 = sin 20. From the latter equation, we have sin 0 = 2 sin 0 cos 0, from which we find that either sin 0 = 0 or
cos 0 = \.
If cos 0 = then cos 20 = 2 cos2 0 - 1 = so cos 0 = cos 20 is not satisfied.
If sin 0 = 0, then 0 = kn for some integer k. If k is odd, then cos 0 = -1 and cos 20 = 1, so we cannot have
cos 0 = cos 20. If k is even, then cos 0 = cos 20 = 1. So, we have cis 0 = cis 20 for even multiples of n, as before.
7.36 By de Moivre's Theorem, we have
(cos 0 + i sin 0)75 = cis750 = cis 750.
Since | cis 750| = 1, if cis 750 is a real number, then it must be equal to 1 or -1.
But cis 750 is equal to 1 or -1 if and only if 750 is a multiple of 180°, i.e. 750 = 180°k for some integer k. Then
_ 180°k _ 12°k
e-~5
Thus, for 0 to be an integer (in degrees), k must be a multiple of 5. Let k = 51. Then
12° • 5t
0 = = 12°f.
The angles 0 of this form in the range 0° < 0 < 90° are 0°, 12°, 24°, 36°, 48°, 60°, 72°, and 84°
7.37
(a) We have r = ^242 + (-24)2 = 24 V2. We seek an angle 0 such that 24 yfl cos 0 = 24 and 24 V2 sin 0 = -24,
which means cos 0 = ^ and sin 0 = - ^. We can take 0 = 7-f, so 24 - 24* = 24 V2e7rn/4
133
_ /64 _ 8 V3
(b) We have r - y (^) + 42 = . We seek an angle 0 such that ^ cos 0 = ^ and ^ sin 0 = 4,
8 V3 _m73
which means cos 6 = 1 and sin 6 = 4^. We can take 0 = ?, so ^ + 4z = 3 e
7.38
(a) e37n/2 = cos y + i sin ^ = | -i |.
(b) 6e7m/6 = 6 (cos 7-f + i sin ) =6 (-4^ - |z) = -3 V3 - 3z
7.39 We see that = cos(-0) + i sin(-0). But
cos 0 - i sin 0 cos 0 - i sin 0

e~w = — = = cos 0 - i sin 0.
0ie
e'° cos 0 + i sin 0 (cos 0 + i sin 0)(cos 0 - i sin 0) cos2 0 + sin2 0
Equating the real parts and equating the imaginary parts, we find cos(-0) = cos 0 and sin(-0) = - sin 0.
7.40 Let z = re10 = r cos 0 + ir sin 0. Then z = re'6 - r ■ eie - re ,B so
z z re,i6 re,-id
- + - + - \e2,e + e 2,e| = | cos20 + isin 20 + cos(-20) + isin(-20)|
z z re -ie re1w
= | cos 20 + i sin 20 + cos 20 - i sin 20| = 2| cos 20|.
Thus, the given equation becomes 2| cos 20| = V3, which means cos 20 = ^ or cos 20 = - —
2 •
If cos 20 = 4^, then 20 = Inn + f = '12,'41)TT or 20 = Inn + H11 = ■■12”v1--— for some integer n, so 0 = (12”^1)n Gr
(12n+ll)n
12
7n _ (12n+7) 7i
If cos 20 = -4^, then 20 = 2roi + ^ = (12”+5)7T or 26 = 2nn + ^ = for some integer n, so 0 = (12”+5)n Gr
(12n+7)7i
12
rg(12n+l)ro/12, rg(12«+5)ra/12^ re(12n+7)ni/U, an(| rg(12n+ll)m/12 , whgre r js a

Therefore, the solutions are of the form
nonnegative real number and n is an integer. If we relax the restriction that r be nonnegative, then all solutions
are of the form fed2”"1"1)711/12 or re(\2n+5)ni/i2, since we have
re(12n+7)ni/12 _ y ^(12n+l)ra712g6m/12 j _ _rg(Un+l)ni/12 an(j rg(12n+ll)m712 _ f ^(12H+5)m712g6m712^ _ _rg(12n+5)7ii/12
7.41
(a) First, we express - \i in exponential form. We have r = y (-^) + (~|) = \[\ = |. We seek an angle
0 such that \ cos 0 = ^ and \ sin 0 = which means cos 6 - ~ and sin 0 = -1. We can take 0 = so
- if = iellra/6. Then
4® _ i,Y = ('Ie..»V6'|8 _ ie44»/3 = J_ Ti + -

4 4 / V2 7 28 256 l 2 2
We also could have tackled this part by taking 0 = -| in the exponential form of ^ - \i. Then, we have
8
' V5 _ 1 A _ n-nutf _ Ie-4„,/3 _ I>/3 . J_ (J_ . V5,\ _ 1 V3.
4 4 / V2 Z 2* 2» * 256 l 2 2 ~ 512 512
134
Review Problems
(b) First, we express -3 V3-3i and 1 + V3i in exponential form. For -3 a/3-3i, we have r = \J(-3 V3)2 + (-3)2 =
V36 = 6. We seek an angle 0 such that 6 cos 0 = -3 V3 and 6 sin 6 = —3, which means cos 0 = — 4^ and
sin 9 = -i. We can take 9 = so -3 V3 - 3i = 6e7ni/6.
For 1 + V3z, we have r = yl2 + (V3)2 = 2. We seek an angle 9 such that 2 cos 9 = 1 and 2 sin 9 = V3,
which means cos 9 = \ and sin 9 = We can take 9 = §, so 1 + V3z = 2eni/3. Therefore,
(-3 V3 - 3z)8 _ (6e7ni/6)s 68e28m73 f 1 V3.\

-13122 - 13122 V3z
(1 + V3z)6 "" (2c™/3)6 ~ 26e2ni ~ 26244 / 2 _ T'J "
7.42
(a) By de Moivre's Theorem, we have cos 59 + i sin 50 = (cos 0 + i sin 0)5, so
cos 50 + i sin 50 = cos5 0 + 5i cos4 0 sin 0-10 cos3 0 sin2 0 - 10/ cos2 0 sin3 0 + 5 cos 0 sin4 0 + i sin5 0.
This time, we equate the imaginary parts, to find
sin 50 = 5 cos4 0 sin 0 - 10 cos2 0 sin3 0 + sin5 0.
Substituting cos2 0 = 1- sin2 0, we get
5 cos4 0 sin 0-10 cos2 0 sin3 0 + sin5 0 = 5(1 - sin2 0)2 sin 0 - 10(1 - sin2 0) sin3 0 + sin5 0
= 5(1-2 sin2 0 + sin4 0) sin 0 - 10(1 - sin2 0) sin3 0 + sin5 0
= 5 sin 0 - 10 sin3 0 + 5 sin5 0-10 sin3 0 + 10 sin5 0 + sin5 0
= 16 sin5 0-20 sin3 0 + 5 sin 0 .
(b) Suppose that there is a function / such that /(sin 0) = sin 40 for all 0. Taking 0 = f, we get /(sin |) = sin ^p,
or /(-^) = Taking 9 = we get /(sin = sin or /(4p) = 4^. But /(4p) cannot be two different
values, so no such function / exists.
7.43 Let w = e711^3 and z = e2ml3l so wz = ent = -1 and |w| = |z| = 1. Also,
w + z = em'3 + e2™73 = Q + ^ = V3z,
which is not real, so part (a) is not necessarily true. We claim that w + z is either real or pure imaginary.
To show that (b) is true, we write w and z in exponential form. They have the same magnitude, so we can
express them as zv = ret9 and z = rel(^. Therefore, wz = rV(0+(^ = r2(cos(0 + <p) + i sin(0 + This is a real number
if and only if sin(0 + <p) = 0, which is true if and only if 0 + ) = r(cos 0 + cos(kn - 0)) + r(sin 0 + sin(fcn - 0)).
This means that
Re(w + z) = cos 0 + cos(kn - 0) = cos 0 + cos(kn) cos 0 + sm(kn) sin 0 = cos 0 + cos(kn) cos 0,
Im(w + z) = sin 0 + sin(/c7i - 0) = sin 0 + sin(/c7i) cos 0 - cos (kn) sin 0 = sin 0 - cos(kn) sin 0,
where the last step in each line uses the fact that sin kn = 0 for all integers k. Therefore, if k is odd, we have
cos(/c7i) = -1, so the real part of w + z shown above is 0. If k is even, we have cos(kn) = 1, which means the
135
imaginary part of w + z shown above is 0. In either case, we see that iv + z is either real or pure imaginary , as
claimed.
7.44 First, we express -4 a/2 + 4 V2z in exponential form. We have r = \J(-4 V2)2 + (4 y/2)2 = 8. We seek an angle
0 such that 8 cos 0 = -4 V2 and 8 sin 0 = 4 V2, which means cos 9 - -Tp arid sin 9 = . We can take 9 - so
-4 V2 + 4 V2z' = 8e3m'/4 = 23e3ra/4.
We can now write the equation as z3 = 23e3m/4. Letting z = re,e, we have z3^*0 = 23e3m/4, so r = 2. Solving
3z’0 = 3tzz/4 gives us z = 2c7T!/4 as one solution. To find the other two solutions, we note that for any cube root of
unity co, we have (2coemZ4)3 = 23co3e3n^4 = 2?e3n^4, because to3 = 1. Therefore, 2coem^ is a solution to the original
equation for each cube root of unity co. We've already found the solution corresponding to co = 1. The ones
corresponding to co = e2m^3 and co = e4m/3 are 2em/4 . e2ni/3 _ | 2enni/u | and 2eni/4 ■ (e2ni/3)2 = \ 2e19nm
7.45 Solution 1. We have ^ = |ôse-)sine' We rationalize the denominator, to get
1 + cos 9 + i sin 9 _ (1 + cos 9) + i sin 9 (1 - cos 9) + i sin 9

1 - cos 9 - ism9 (1 - cos 9) - isin 9 (1 - cos 9) + isin9
(1 + cos 0)(1 - cos 0) - sin2 0 + z(sin 0 + cos 0 sin 0 + sin 0 - cos 0 sin 0)
(1 - cos 0)2 + sin2 0
1 — cos2 0 - sin2 0 + 2i sin 0 _ 2i sin 0 _ i sin 0
1-2 cos 0 + cos2 0 + sin2 0 2-2 cos 0 1 - cos 0
From the half-angle formula for tangent, we have tan |0 =

— l-cos9
sin g ,
/sing _ ;
!_COS0 - i'-ui „.
t g
l+e,e el'8/2_|_e i8/2

Solution 2. We have ~ Dividing both the numerator and denominator by el9/2, we get 2_g»0 ^-10/2 giQ/2 •
g _ e,eO-e l8^2 . je/2+p-ie/2 2 cos I

itus; • , «
Since cos | = and sin | = 2i , we have hsjihm = -V
2; sin ■
= z cot f.
7.46
(a) Solution 1. The roots of unity are c°, c27T,/”, g4™/", ^ ^(n-i)m/n, so product is es, where
. „ 27tz 4tiz 6ni 2(n - l)ni 2 ni 2ni (n - l)zz

S = 0 + — +-+-+ ••• + —-— [1 + 2 + • • • + (n - 1)] (zr - l)rrz.
n n n n n ~Z7 2
If zr is even, then (n - 1)ti is an odd multiple of n, so es = = -1, and if n is odd, then (n - l)7i is an
even multiple of n, so es = gW1)711 = 1. So to summarize, the product of the nth roots of unity is -1 if n is
even, and 1 if n is odd.
Solution 2. In the text, we proved that if co is an zzth root of unity, then so is Thus, the zith roots of unity
can be paired so that the product of the numbers in each pair is equal to 1. The only roots that cannot be so
paired are 1 and -1, because they are their own reciprocals.
If n is odd, then -1 does not appear among the nth roots of unity, so in this case, the product of the roots
is 1. However, if n is even, then -1 does appear among the zzth roots of unity, so in this case, the product of
the roots is -1.
Solution 3. The nth roots of unity are the roots of the equation xn -1 = 0. Letting r\, r2,...,rn be the roots, we
can write the equation as (x - r\){x - 7^) • • • (x - rn) = 0. In the expansion of the left side, the only constant
term appears when we take the -r, from each factor, giving us the product (-l^zqz^ • • • rn. Since the constant
term in x" - 1 is -1, we have (-l)"z'iZ'2 • • • rn = -1. Therefore, the product of the roots is (-1)1-". Hence, the
product of the nth roots of unity is -1 if n is even, and 1 if n is odd. (This approach is essentially equivalent
to letting x = 0 in the factorization xn - 1 = (x — r\)(x - r?) ■ • • (x - rn).)
136
Review Problems
(b) We take a cue from Solution 2 of part (a). We know that the nth root of unity e2kn^n is primitive if and only
if gcd(k,n) = 1. But
oo,2nni/n - a,2(n-k)ni/n
0j2kni/n Îkni/n
and if gcd(k, n) = 1, then gcd(n - k,n) = 1, so tt, is also a primitive nth root of unity. Thus, the primitive
nth roots of unity can be paired so that the product of the numbers in each pair is equal to 1, and again, the
only exceptions are 1 and -1.
We see that 1 is a primitive nth root unity only for n = 1, and -1 is a primitive nth root of unity only for
n -2. Hence, the product of the primitive nth roots of unity is 1 for all n ^ 2, and -1 for n = 2.
7.47
(a) Let X\ be a root of x2"+1 + 1, so xf,+1 + 1=0. Let x2 = -X\. Then
.2(1+1 2n+l ,2n+l

1 = -x2n+1 - 1 = -(xf +1 + 1) = 0,
~ 1 = (-*l)
so x2 is a root of x2,,+1 — 1. Hence, the roots of x2n+l + 1 are the negatives of the roots of x2"+1 - 1.
(b) If x2 = —X\, then geometrically, x2 is the reflection of x\ through the
origin in the complex plane. (Alternatively, we can say that x2 is a 180°
rotation of X\ about the origin.) In addition, the roots of x2"+1 - 1 form
the vertices of a regular (2n + l)-gon, as do the roots of x2n+1 + 1. Since
(x2«+i _ i)(x2d+i -(-!) = x4n+2 - 1, the two sets of roots together form the
vertices of a regular (An + 2)-gon.
As an example, we plot the roots for n = 2. The roots of x5 - 1 are
marked by a dot, and the roots of x5 + 1 are marked by a diamond.
7.48 Let w = re'6 in exponential form. Let zjt = rV»e(S+2kn)t/nf for i <k <n. Then
zn _ [rl/ne(Q+2kn)i/nyi _ re(6+2kn)i _ rei0 . ^2kni _ w
Hence, Z\, z2,..., zn are the n roots of z" = iv. These roots all have the same magnitude, and their arguments are
—^, e+„nn, which form an arithmetic sequence with common difference Therefore, the roots form
the vertices of a regular n-gon.
7.49 First, we express 1 + V3z in exponential form. We have r = i/l2 + (V3)2 = 2. We seek an angle 6 such that
2cos 9 = 1 and 2 sin 6 = V3, which means cos 6 = \ and sin 9 = We can take 9 = so 1 + V3z = 2em/3.
Then x19 + z'yi9 = (2eni/3)19 and x9i + z'y9i = (2em/3)91. We could use these equations to find xi9,yi9,x9i, and y9i
individually. But instead, we'll show a slicker solution. We don't need to find each of these values; we are only
asked for xi9y9i + x9iyi9. This expression will appear if we take the product of the equations xi9 + zyi9 = (2em/3)19
and x9i + zy9i = (2eni/3)91. This product gives us
(xi9 + zyi9)(x9i + zy9i) = (2enl/3)U0.
Expanding the left side gives us
(xi9 + zyi9)(x9i + zy9i) = xi9x9i + zxi9y9i + zx9iyi9 + i2yi9y9i = (î9^9i - yi9y9i) + Kxi9})n + X9\V\9)-
Thus, Xi9y9i + x9iyi9 is the imaginary part of (2era/3)110 = 2110g110ra/3 = 2110 (-\ + ^z) = -2109 + 2109 V3z, so the
answer is k 109
7.50 Let z = x + yz. Then e2 = ^+y! = e* ■ = ^(cos y + i sin y), so ez = H(cos y - i sin y). Also, e2 = e* ^ = fe v1 =
^(cos(-y) + isin(-y)) = ^(cosy - zsiny). Hence, ez = e2.
137
7.51
ex + e ex - e + 2 + e-2* e2x-2 + e~2x

(a) cosh2 x - sinh2 x = = 1.
_ et+y-e-x-y
(b) We have sinh(x + y) , and
px — p x (>y + g y gx + g x ey - e y
sinh x cosh y + cosh x sinh y - —-— • —--1--— • ———
gx+y + gX-y _ e~x+y _ e-x-y gx+y _ gx-y + e-x+y _ e-x-y
2ex+y - 2e~x~y ex+y - e~x~y
Hence, sinh(x + y) = sinh x cosh y + cosh x sinh y.

eX-rx
(c) _ sinh* so tanh(x +

In general, we have tanhx = ^ = ^ = j. iA
y) =- cosh{x+y)
smh^+y) ^Ve ^ye an identity for
sinh(x + y); let's see if we can make one for cosh(x + y). We note that the identity for sinh(x + y) looks a
lot like the identity for sin(x + y). Maybe the identity for cosh(x + y) will look a lot like the identity for
cos(x + y). Let's write cosh(x + y), cosh x cosh y, and sinh x sinh y and see if we can relate them:
gX+y + g~{x+y)
cosh(x + y) =---,
e* + e~x ey + e~v _ ex+v + ex~y + eV~x + e~{x+y]

cosh x cosh y =
2 2 _ 4 '
gX _ g-x ey _ g-y gX+y _ ^t-y _ £y-x + g-(x+y)
sinh x sinh y =
2 2 4
Sure enough, when we add cosh x cosh y and sinh x sinh y, we get cosh(x + y). So, we have
sinh. x i sinh y
sinh(x + y) _ sinhxcoshy + coshxsinhy _ + coshy _ tanhx + tanhy
tanh(x + y) =
cosh(x + y) cosh x cosh y + sinh x sinh y \ + sinh.vsir.h:/ i + tanh x tanh y
(d) We have sinhx + sinh y = ^ e x+^y e y, and

e(x+y)/2 _ g-{x+y)l2 e(x-y)/2 + g-(x-y)/2
e* + ey -
2 sinh ^ + y cosh x-y = 2-
2 2 2 2 2
Hence, sinh x + sinh y - 2 sinh ^ cosh .
7.52 Let f(n) = xn + yn. First, we express x and y in exponential form. We see that x = i - cos ^ +
zsin ^ = e2™13, and y = = cos y + zsin^1 = e4ra/3. Then x3 = e2™ = 1 and y3 = c4m = 1. Hence,
f(n + 3) = x"+3 + yn+3 = x" • x3 + y” • y3 - xn + yn = f(n). This tells us that f(n) = /(0) if n is divisible by 3, f(n) = /(l)
if ft is one more than a multiple of 3, and /(ft) = /(2) if ft is two more than a multiple of 3. Therefore, we only need
to check /(0), /(l), and /(2).
We see that /(0) = x° + y° = 2, and /(l) = x + y = -1. To compute /(2), note that x2 = e4™/3 = y. Therefore,
/(2) = x2 + y2 = y + x4 = y + x = /(1) = -1. Hence, /(ft) = -1 if and only if ft is not a multiple of 3
Challenge Problems
7.53 The polynomial z9 - 1 factors as z9 - 1 = (z3 - l)(z6 + z3 + 1). Hence, the roots of z6 + z3 + 1 are the ninth
roots of unity that are not cube roots of unity.
138
Challenge Problems
The ninth roots of unity are cisO°, cis 40°, cis80°,..., cis320°. The cube roots of unity are cisO°, cis 120°, and
cis 240°. The only ninth root of unity that is not a cube root of unity, and whose argument is between 90° and 180°,
is cis 160°, so the answer is 160°
7.54 Since x5 -1 - (x- l)(x4 + x3 + x2 + x + 1), and z t 1 , z must satisfy the equation z4 + z3 + z2 + z + 1 = 0. Then
,16 , „17 , „18 , „19_15/i , _ . „2 , „3 _4\ _
z15 + z16 + z17 + z18+z19—15'
zI5(l +Z + ZZ+ Z3 +Z4) = 1-0 = 0.
Similarly, each 5 consecutive terms in the sum add to 0, leaving only z50 = |~T].
7.55 We factor the first two terms and the second two terms to write the equation as z4(z3 + 8) - 2i(z3 + 8) = 0.
Aha! We can factor more, to get (z4 - 2z)(z3 + 8) = 0. Therefore, the solutions to the equation are the solutions of
z4 = 2i and z3 = -8.
Since 2i - 2eml1, the solutions to z4 = 2i are the four numbers of the form 2x,4<jjeml8, where co is a fourth root of
unity. The fourth roots of unity are 1,eni/2 em, pe3n,/2.
3nil2 So,
^ the four solutions to z4 =
- 2z are given by the four complex
numbers Vie71'18, y/2e5ni/8, \/2e9ni/8, y/2e13ni/8
Since -8 = 8ent, the solutions to z3 = -8 are the three numbers of the form 2coen^3, where a; is a cube root of
unity. The cube roots of unity are l,e2ni,3,e4ni/3, so the solutions to z3 = -8 are 2enil3,2eni, 2e5ni/3
7.56 The argument of the complex number z = 1 + 2i is 0. Then
z2 = (1 + 2 i)2 = 1 + 4z - 4 = -3 + 4 i,
z4 = (z2)2 = (-3 + 4z')2 = 9 - 24z - 16 = -7 - 24z,
z8 = (z4)2 = (-7 - 24z)2 = 49 + 336z - 576 = -527 + 336z.
Since the real part of z8 is negative and the imaginary part is positive, 80 lies in the second quadrant
7.57 We can rewrite the given equation as z9 + z6 + z3 + 1 = 0. Since z12 - 1 factors as
z}2 — 1 — (z3 - l)(z9 + z6 + z3 + 1),
the roots of z9 + z6 + z3 + 1 = 0 are the 12th roots of unity that are not cube roots of unity.
The 12th roots of unity are e°, e2ml12, e4m/12, ..., e21ml12. The cube roots of unity are e°, e2ni/3 - e8m/n, and
e4m/3 _ gi6m/i2_ xherefore, the roots of z9 + z6 4- z3 + 1 = 0 are
g2m/12 g4ra'/12 g6ra712 g10ni/12 eV2mlV2 eUni/12 g18ni/12 g2pro/12 and ^2ni/U
7.58 Since (x - l)(x2 + x + 1) = x3 - 1, the roots of x2 + x + 1 = 0 are the cube roots of unity other than 1. Let these
roots be a and
Let fn(x) = x2n + 1 + (x + l)2n. Then by the Factor Theorem, fn(x) is divisible by x2 + x + 1 = (x - a)(x - (1) if and
only if fn(a) - /„(/?) = 0. We see that /„(«) = a2n + 1 + (a + l)2". Since a2 + a + 1 = 0, we have a2 - -a - 1, so
(a + l)2n = ((a + l)2)n = (a2 + 2a + 1)” = (-a - 1 + 2a + l)n = a”.
so
fn(a) = a2n + 1 + (a + l)2” = a2n + 1 + an = a2n + an + 1.
This is a familiar form; if we let y = a", we have y2 + y + 1, which equals ^ if y * 1. So, if an 11, we have
a3n - 1
Met) =
an - 1 ’
139
If an = 1, we have /„(«) = a2n + an + 1 = 1 + 1 + 1 = 3, so a is not a root of /„(«) if an = 1. Since n is a primitive

cube root of unity, we have an = 1 if and only if n is a multiple of 3.
If n is not a multiple of 3, then an t 1, and fn(a) = However, because a is a cube root of unity, we
have a3n - (a3)n = 1" = 1, so /„(«) = 0, which means a is a root of fn(a). Notice that nothing in this argument
depends on which primitive cube root of unity a is, so we have shown that both roots of x2 + x + 1 are roots of
x2n + 1 + (x + \)2n if and only if n is not a multiple of 3
7.59 Let zo - 2z, so z = zo/2. Then
32z5 + 16z4 + 8z3 + 4z2 + 2z + 1 = (2z)5 + (2z)4 + (2 z)3 + (2z)2 + 2z + 1 = w5 + w4 + zo3 + w2 + w + 1.
Since zo6 - 1 = (zo - 1 )(io5 + iv4 + zo3 + w2 + w + 1), the roots of zo5 + zo4 + zo3 + zo2 + zo + 1 are the sixth roots of unity
lp2ni/6 1Ani/6 lp6ni/6 \ 8ni/6 i 1 Wni/6
other than 1. Hence, the solutions in z are ^ ^ C ^ ^ ^ C j Cli Ivt r\ C
7.60 A complex number z is real if and only if z = z. Since x is a root of unity, we have |x| = 1, so r = 1/x.
Similarly, y = l/y. Then
x+y
(*+»)> = (*+?)*= g + i)‘=( xy {-+ yf- = ^X
xkyk
(x ++ y)
y)k'
where xk = 1 and yk = 1 because x and y are kth roots of unity. Since (x + y)k - {x + yf, we know that (x + yf is
real. (See if you can find a geometric approach to solve the problem!)
7.61 To evaluate the given sum, we pair the terms. Let 1 < k < 1996. Then
1 1 1 + (01997~k + l+0)k 2 + cok + co 1997-k 2 + of + CO 1997-k

+ = 1.
1 + COk 1 + C01997~k (1 + COk)( 1 + C01997~k) 1 + COk + CO1997 k + Ol1997 2 + COk + CO1997 k
Hence,
+
1 +00 1 + Ol1996
1 1
+
1 + CO2 1 + CO1995
1 1
1 + CO3 1 + CO
1994 = 1,
1 1
1 + co998
+
1 + a;999
= 1,
and
1+a;1997 1+1 2'

Therefore,
1 1 1997
+ T-^ + ••• +
1+01 1 + 012 1 + ai1997 998 + 2
7.62 Note that |z,n+11 — = = 1 for all n > 0. We are given that |zq| = 1, so |z„| = 1 for all n > 0. Then
z„z„ = |z„|2 = 1, so zn = l/z„. Hence,
_ lz» _ lZn . 2
z«+1 — _ —
Zn
140
Challenge Problems
for all n> 0.
Then
Zl — ZZg,
,2
Z2 = iz\ = Z’(z'Zg)2 = -ZZg,
Z3 = ZZ2 = z(-ZZq)2 = —ZZg,
Z4 = ZZ3 = z'( —ZZg)2 = —ZZg6,
and so forth. In general, we see that z„ = -ZZg" for all n >2. Hence, the equation Z2005 = 1 becomes
22005
-ZZn 1,
,2005
or Zg'01 = i. This equation has 22005 roots, so there are possible values for zg.
7.63 The recursive definitions we are given for an+1 and bn+1 have the form of the angle sum and angle difference
identities for sine and cosine. We consider taking advantage of this by letting a and jS be a sine or cosine, but then
we note that a and p can be greater than 1 or less than -1. But thinking of sines and cosines leads us to complex
numbers, which also produce expressions of these forms.
We let z = a + pi, and let z„ = an + bni. Then
zz„ = (a + pi)(an + bni) = aan + abni + fiani - pbn = aan - fibn + (p>an + abn)i = an+\ + bn+\i - zn+\.
Also, zi = a + pi = z, so zn = z" for all positive integers n.
If #1997 = #i and &1997 = b\, then Z1997 = Zi. Since zM = z" for all integers n, we must have z1997 = z. Rearranging
gives z1997 - z = 0, which factors as z(z1996 — 1) = 0. Hence, there are 1997 possible values of z = a + pi, namely 0
and the 1996th roots of unity, which means there are 1997 possible pairs (a, p).
7.64 By de Moivre's Theorem, we have cos nO + z sin nd = (cos d + i sin 9)n. But by the Binomial Theorem,
n n-1 n
(cos6 + isin9)n - cos" 9 + i\ I cos 0sin0 cos" 2 9 sin2 9 - ( ) z cos" 3 sin3 9 + • • • .
Equating the real parts and equating the imaginary parts, we get
n n-2
cos n9 = cos" 0 - | | cos 9 sin2 0 H-,
sin 770 = ( ” j cos" 1 0 sin 0 - f ” ) cos" 3 sin3 0 +
Now, we can tackle tan n9:
sin n9 (?) cos" 1 0 sin 0 - (?) cos" 3 sin3 0 +

tanzi0 =-— =-——~7zr.-, . 2 o-
cos n9 cos” 0 - (?) cos”-2 0 sin2 0 h-
Dividing the numerator and denominator by cos" 0, we find
(?) cos- e sin e - a cos”-3 sin3 6 + ■•■ (?) gj - Q ;3 0 + (?) tan 0 - (?) tan3 0 +
cos" 6 - (”) cos"-2 6sin2 6+ ■■■ j _ g)ania + 1 - (?) tan2 0 + • • •
cos2 8
7.65 From the given equation, we get z2 - (2 cos 3°)z + 1 = 0. Then by the quadratic formula, we have
2cos3° ± V4cos23° - 4 2cos3° ± y/-4(1 - cos23°) _ 2cos3° ± 2z Vsin23° _ QO

z = -------— — cos d x / sm d .
141
By de Moivre's Theorem,
1 V3
(cos3° ± zsin3°)2000 = cos6000° ± isin6000° = cos240° ± isin240° = -- T -^-z,
so z2000 + ^25oo = ~\ ~ “ 2 + = The êast integer that is greater than -1 is [o].
7.66
(a) Neither x" - 1 nor any of the O^(x) has repeated roots, and all of these polynomials are monic. Also, no
complex number co can be the root of two different O^.(x), since, by definition, a; is a root of (x) if and
only if di is the smallest positive integer such that codi = 1. Therefore, if we show that each root of x" - 1 is a
root of exactly one of the 0</;(x), and that each root of each O^.(x) is a root of x” - 1, then we can conclude
that x" - 1 equals the product of the O^(x).
First, let wbea root of ^.(x), so codi = 1. Since di is a divisor of n, we have k - f for some integer k.
Therefore, we have con = (g/1) / = (<ud') = lk = 1, so co is a root of x" - 1. This tells us that every root of
each Orf, (x) is a root of xn - 1.
Next, we let a; be a root of xn - 1, so con = 1. The set of positive integers k such that cok = 1 is non-empty.
Let d be the smallest such positive integer. Then a) is a primitive dth root of unity, so a)d = 1.
Let q and r be the quotient and remainder when n is divided by d, respectively, so n = qd + r and
0 < r < d — 1. Then con = coCfd+r = 1. But co^d+r - (a/)'? • g/ = a/, so cor = 1. But r < d — 1, so r cannot be
positive. (By definition, d is the smallest positive integer such that a/ = 1.) Therefore, r = 0, which means
n = qd. Hence, d must be a divisor of n. This tells us that every root of x" - 1 is a root of of some Od,(x).
We conclude that x" - 1 can be factored as (x)<f>d2 (x)Od3(x) • • • (x), where d\, d2/ d2,...rdk are all of
the positive divisors of n.
(b) Taking n = 1 in the identity in part (a), we get Oi(x) = | x - 1 .

Taking n = 2, we get Oi(x)02(x) - x2 - 1, so
x2 - 1 (x - l)(x + 1)
02(x) = x+1
<E>i (x) x-1
Taking n = 3, we get Oi(x)3(x) = x3 - 1, so
. x3 - 1 (x - l)(x2 + X + 1)
x2 + X + 1
= m =-T=l-
Taking n - 4, we get <J>i(x)02(x)<l>4(x) = x4 - 1, so
x4-l
04(x) =
O] (x)02(x)'
But Oi(x)02(x) = x2 - 1, so
x4 - 1 (x2 - l)(x2 + 1)
04(x) = x2 + 1
x2 - 1 X2 - 1
Taking n = 5, we get Oi(x)Os(x) = x5 - 1, so
x5 - 1 (x - l)(x4 + x3 + x2 + x + 1)
05(x) = x4 + X3 + X2 + X + 1
<fh(x) X -1
142
Challenge Problems
Taking n = 6, we get Oi(x)02(x)<E>3(x)6(x) = x6 - 1, so
, , . x6 - 1
®i(x)02(x)03(x)'
But Oi(x)<E>3(x) = x3 - 1, 02(x) = x + 1, and the numerator factors as
x6 - 1 = (x3 - l)(x3 + 1) = (x3 - l)(x + l)(x2 - x + 1),
so
(x3 - l)(x + l)(x2 - X + 1)
ch(x) = x2 - x + 1
(x3 - l)(x + 1)
Taking n = 7, we get 0j(x)07(x) = x7 - 1, so
_ . . x7 - 1 (x - l)(x6 + X5 + X4 + X3 + X2 + X + 1)
x6 + x5 + x4 + x3 + x2 + X + 1
=-131-
Taking n = 8, we get Oi(x)2(x)<E>4(x)8(x) = x8 - 1, so
x8 — 1
08(x) =
0i(x)02(x)04(x)
But 0i(x)02(x)04(x) = x4 - 1, so
x8 - 1 (x4 - l)(x4 + 1)
08(x) = x4 + l
x4 - 1 x4 - 1
Taking n = 9, we get i(x)3(x)<t>9(x) = x9 - 1, so
x9 - 1
09(x) =
<E>i(x)03(x)‘
But Oi(x)03(x) = x3 - 1, so
x9 - 1 (x3 - l)(x6 + x3 + 1)
09(x) = x6 + x3 + 1
x3 - 1 x3 - 1
Taking n = 10, we get Oi(x)<E>2(x)05(x)<J>io(x) = x10 - 1, so
x10-l
°lo(x) 01(x)02(x)05(x)’
But Oi(x)05(x) = x5 - 1, 02(x) = x + 1, and the numerator factors as
x10 - 1 = (x5 - l)(x5 + 1) = (x5 - l)(x + l)(x4 - x3 + x2 -X + 1),
so
(x5 - l)(x + l)(x4 - X3 + x2 - X + 1)
<Pio(x) -
x4 - x3 + x2 - X+ 1
(x5 - l)(x + 1)
(c) Taking n = p, we get <t>i(x)Op(x) = xp - 1, so
xP-1 xP-1
%(x) = XP 1 + xp 2 + • • • + X + 1
Oi(x) x—1
143
(d) Let co - e2n^n, and let cok be a primitive nth root of unity. Then, we have gcd(n, k) = 1, so gcd(n, n-k) = 1 as
well, which means con~k is a primitive root of unity. Since n > 2, we know that we do not have k - |, so cok
and con~k are distinct. Therefore, the roots of „(x) come in pairs (and none is repeated), so the degree of
On(x) is even.
(e) Since the degree of <P„(x) is even, the leading coefficient of <t>„(-x) is 1. Also, neither O2„(x) nor Om(-x) has
any repeated roots. Therefore, to show that 02«(x) = 0„(-x), it suffices to show that the roots of <1>2n(*) and
<£>„(-x) are the same. So, we will show that every root of On(-x) is a root of 2„(x), and vice versa.
Every root of 0„(-x) is the negative of a root of 0„(x). The roots of <f>„(x) are the primitive nth roots of
unity, so each is of the form elnkl^n, where k is an integer such that gcd(k, n) = 1. We must therefore show
that -e2nkl/n is a primitive (2n)th root of unity. Since -1 = em, we have -e2nkl/n = eme2nkl/n - e(2k+n)™/n. We
then have
,2 n
^_e2nki/n g(2k+n)ni/n ,2(2k+n)m
=e = h
so g(2J:+n)7I!/'! is a (2^1)* root of unity. But is it a primitive (2n)th root of unity? To see that it is, we note that
e(2k+n)m/n _ e2(2k+n)m/{2n) _ ^2m/(2n)^2fc+n_ gQ/ we must show that gcd(2ft: + n, 2n) = 1. Since n is odd, we know
that 2k + n is odd. Since gcd(k, n) = 1, there are no divisors of n (besides 1) that are also divisors of 2k + n.
So, we have gcd(2k + n, 2n) - 1, which means that (e2mIQ-n)fkJrn incjeeci a primitive (2n)th root of unity.
Going the other direction, we must show that every root of 02„(x) is also a root of 0„(-x). To do so,
we must show that the negative of any primitive (2n)th root of unity is a primitive nth root of unity. Let
co = g27*/!2") be a primitive (2n)th root of unity, so gcd(/c,2n) = 1. Specifically, k is odd. We then have
_ _g2nki/(2n) _ ^ni^2nki/(2n) _ g(2k+2ri)ni/(2n) _ g(k+ri)ni/n
This gives US (-ce)n = (e(k+n)m/ny _ e(k+n)m' an(J n are odd, the sum k + n is even, which means
(-co)n = e(k+n)m = 1, so -co is an nth root of unity. To show that it is a primitive nth root of unity, we note
that g(fc+”)ra/” = ^g2ni/n^(k+n)/2^ so we must show that gcd = 1- (Remember, k + n is even, so — is an
integer.) Since gcd(k,2n) = 1, we know that gcd (k,n) = 1 as well. Therefore, we have gcd (A: + n,n) = 1, so
gcd = 1, as desired.
We have therefore shown that every root 0„(-x) is a root of 02„(x), and vice versa, which means 02);(x)
and 0„(-x) are the same.
7.67 The expressions on the left sides of the desired equations resemble applications of the Binomial Theorem,
which states that
n _ ji-1, n-2.,2 n
(x + y) xny° + y + X y + • • • +
n- 1
xy-1 + x°yn.
We can substitute any numbers for x and y. We use the desired equations to guide us. We'd like the Q) term to
have a coefficient of -1, but the powers of x and y for this term in the Binomial Theorem are both even. So, we
use complex numbers! We let x = 1 and y = i We then have
Thus, the sums in the problem represent the real and imaginary parts of (1 + if.
In exponential notation, we have 1 + i = Vê71^4, so
(1 + if = (yfleni,A)n = (V2)Vra/4 = 2n/2 (cos ~ + ism = 2n/2 cos — + i2n/2 sin —.

\4 4/ 4 4
144
Challenge Problems
Equating the real parts and equating the imaginary parts gives us
7.68 Let S = w + 2w2 + 3w3 + • • • + 9w9. We can produce another series with many of the same powers of w by
multiplying both sides of this equation by zv. (We might be inspired to do this by the standard proof of the formula
for sum of a geometric series.) We find zvS = w2 + 2w3 + 3zv4 + ■ • ■ + 9zv10. Subtracting this from the equation for S
gives us a series we know how to handle:
(1 - w)S = iv + w2 + w3 + ■ ■ ■ + w9 - 9zvw = w( 1 + zv + w2 +...w8) - 9ww.
Since w9 - (cos 40° + z'sin40°)9 = cos 360° + i sin 360° = 1, we know that zv is a ninth root of unity besides 1.
Therefore, \ + iv + w2 -hw8 = 0 and zvw = w9 -w = zv, which means we have S(1 - zv) = -9zv from the equation
above. Therefore, S = -9w/(l - w), and
1 1 - iv \W — 1| |zt7 - 1|
(7.1)
s 9w \9w\ 9
since \w\ - 1. Writing w as cos 40° + i sin 40°, we have
\w - 1| = (cos 40° - l)2 + (sin40°)2 = \J(cos2 40° + sin2 40°) + 1- 2 cos 40° = V2 - 2 cos 40°.
However, we have to answer in the form a sin b. To do so, we recognize that V2 - 2 cos 40° looks a lot like the
half-angle formula for sine:
. 6 1 - cos 6
sm- =
Letting Q = 40°, we have sin 20° Vl - cos 40°, so V2 - 2 cos 40° = 2 sin 20°, which means that |S| 1 =
After solving Problem 7.75, come back to this one and see if you can find a faster way to finish starting from
line (7.1).
7.69 We prove the result by induction on n. The result clearly holds forn = 0 and n = 1, so assume it holds for
n - k and n = k +1, for some nonnegative integer k, which means Tjt(cos 0) = cos k6 and Tfc+i(cos 0) = cos((k+1)0).
Then
Tk+2(cos 0) = 2 cos 0 • Tjt+i(cos 0) - Tk(cos 0)

= 2 cos 0 cos((k + 1)0) - cos kd.
We want to show that this is equal to cos((k + 2)0).
By the sum-to-product formula for cosine,
cos((k + 2)0) + cos kd -2 cos((k + 1)0) cos 0,
so 2 cos((k + 1)0) cos 0 - cos kd = cos((fc + 2)0). Therefore, Tk+2(cos 0) = cos((k + 2)0).
Hence, the result holds for n = k + 1 and n = k + 2, and by induction, it holds for all nonnegative integer n.
145
7.70 Consider the polynomial with e'a, elP, and eiy as roots:
(z - eia)(z - eV)(z - eiy) = 0.
We think of this polynomial for a couple reasons. First, it's a cubic, which we might be able to use to make terms of
the form e3ia. Second, we know that when we expand the product, the coefficient of the z2 term will be 0 because
the sum of the roots is 0. Specifically, expanding the product gives
z3 — (eia + elP + eiy)z2 + (emeiiP + eiaeiy + eêiy)z — eiae'Peiy = 0.
Now, we just need to figure out what the coefficient of the z term equals. We can get something close by multiplying
ela + elP + ety = 0 by e~l0le~lPe~ty, which gives us e~me~lP + e~lae~iy + e~'Pe~ly - 0. If only we could turn those negative
exponents positive... But we can! We take the conjugate of both sides of this equation. On the right, we have 0,
and on the left, we have
e-iae-'P + e-ê-'y + e-'Pe-v = triae-'P + eê^y + e^Pe^y

= eiajP + eiaeiy + e'VC
So, we have eiaelP + e'aety + elPely = 0. (We could have gotten here a little faster with a little more insight. Taking
the conjugate of the equation ela + elP + e'r = 0, we get e~m + e~TP + e~ly = 0. Multiplying both sides by eiaelPeiy, we
get eiae‘P + eiaeiy + && = 0.)
We now know that the coefficients of z2 and z are 0, so our polynomial is z3 - eme1Peiy = 0. Taking z = eia, we
find e3ia = eiae‘Peiy. Similarly, taking z = e'P and z = eiy, we find e3lP = eiaelPeiy and e3iy = emelPeiy. It follows that
g3ia _ g3;/S _ ^3iy _
7.71 We have the identity cos 9 = e'e+£ . Let a = elxl1, b = eiy/2, and c = e!z/2. Then cos x = e'*+2e " = “2+2/a'.
Similarly,
b2 + l/b2 c2 + l/c2 a2b2c2 + l/(a2bzc2)
cosy =--- cosz=---, cos(x + y + z) =---,
x + y\ _ ab + 1 /(ab) x + z\ ac + 1 / (ac) y + z\ _ be + 1 /(be)

cos cos cos
m= m
Therefore,
a2 + l/a2 b2 + l/b2 c2 + l/c2 a2b2c2 + l/(a2b2c2)

cos x + cos y + cos z + cos(x + y + z) =---h---1---h-—-1
111
a2b2c2 +az + b2 + c2 + ^r + T^ + -^ +
2 b2 c2 a2b2c2 J '
and
„ /x + u\ (x + z\ /y + z\ „ ab + l/(ab) ac + l/(ac) bc + l/(bc)
4cos ( V) cos (-2-) cos (»—) = 4 ---^-
1
ab + ac + be +
2 ab ac be
1 ,
a2b2c2 + a2 + b2.2 + c2
,2 + ^ + ~1
+
1
2 a2 b2 c2 a2b2c2 J ’
Hence,
cosx + cos y + cosz + cos(x + y + z) = 4cos cos (—y—) cos + Z) .
We also have the identity sin 6 = e'e so
a2-1/a2 . b2-l/b2 . c2-l/c2 . a2b2c2 — l/(a2b2c2)

sinx=-—-, siny=-—-, smz =-—-, sm(x + y + z) =-
2i 2i 2i 2i
146
Challenge Problems
x +y\ _ ab- l/(ab) x + z\ ac-l/(ac) y + z \ _ be - 1 /(be)

sin sin
2i m= 2i 2/
Therefore,
. . , . a2— l/a2 b2 — l/b2 c2 — lie2 a2b2c2 — l/(a2b2c2)

sin x + sin y + sm z - sin(x + y + z) = -— H--— + v
2i 2i 2i 2i
1 1 1
= ~ ( -a2b2c2 + a2 + b2 + c2 - A- — +
2i a2 b2 c2 a2b2c2 J'
and
4 sin sin ) sin f ) = 4-a^ ~ ~ ^/(ac) . be - l/(be)

\ 2 J \ 2 J \ 2 J 2i 2i 2i
= Xiab ac - be -
2i3 ab ac be
j_ 1
2i
a2b2c2 -a2-b2-c2 + Az + A^ + A^-
a2 b2 c2 a2b2c2
1 1 1 1 1 \
Hence, sinx + siny + sinz - sin(x + y + z) = 4sin ^ (~Y“) S4n
7.72 If 0 = y, then cos 76 = cos2ti = 1. In an Exercise in Section 7.1, we proved the identity
64 cos7 0-112 cos5 0 + 56 cos3 0-7 cos 0 = cos 70,
so for 0 = y, we get
,, 7 271 „„„ s 2n ^ ,271 „ 271
64 cos7 —— 112 cos5 — + 56 cos3 —— 7 cos — = 1.
7 7 7 7
Thus, x = cos y is a root of the equation 64x7 - 112x5 + 56x3 - 7x — 1 = 0. But 0 = y and 0 = y also satisfy
cos 70 = 1, so x = cos y and x = cos y are also roots of the equation 64x7 - 112x5 + 56x3 - 7x - 1 = 0. We have a
polynomial of degree 7, and we know three of the roots, so we need to think about what the other four roots are.
We go back to the equation

cos 70 = 1.
In addition to 0 = y, y, and y, we see that the other solutions in the interval [0,27i) are 0, y, y1, and y5,
and that there are no other values of 0 in the interval [0,2n) that satisfy the equation. We can quickly see that
x = cos 0 = 1 is a root of 64x7 - 112x5 + 56x3 — 7x — 1 = 0. So, we know that cos 0, cos y, cos y, and cos y are roots
of 64x7 - 112x5 + 56x3 - 7x - 1 = 0, but what are the other three roots?
We found the polynomial 64x7 - 112x5 + 56x3 - 7x - 1 = 0 by starting with the equation cos 70 = 1, which has
7 solutions for 0 6 [0,27i), namely, 0, y, y, y, y, y5, and y1. We've seen that the cosines of the first four are
roots of 64x7 - 112x5 + 56x3 - 7x - 1 = 0. Noting that cos y = cos y, cos y1 = cos y, and cos y1 = cos y, we
conjecture that the polynomial has three double roots corresponding to these equal cosines.
Let's see if that guess is correct. We know that one of the factors of the polynomial is x - 1, so we take out a
factor of x - 1 to get
64x7 - 112x5 + 56x3 - 7x - 1 = (x - l)(64x6 + 64x5 - 48x4 - 48x3 + 8x2 + 8x + 1).
If our guess that the polynomial of degree 6 is the square of a cubic is correct, we must have
64x6 + 64x5 - 48x4 - 48x3 + 8X2 + 8x + 1 = (ax3 + bx2 + ex + d)2.
147
If we were to expand the righthand side, we would get
64x6 + 64x5 -48x4 -48x3 + 8x2 + 8x 4- 1 = (ax3 + bx2 + cx + d)2 = a2x6 + 2abx5 + ■■■ .
Equating the corresponding coefficients of x6 and x5, we get a2 - 64 and lab = 64. Without loss of generality, we
can take a = 8. (If a = -8, then we can flip the sign of all the coefficients of the cubic.) Then 16b = 64, so b = 4.
Hence,
64x6 + 64x5 - 48x4 - 48x3 + 8x2 + 8x + 1 = (8x3 + 4x2 + cx + d)2.
Again, expanding the right-hand side, we get
64x6 + 64x5 - 48x4 - 48x3 + 8x2 + 8x + 1 = 64x6 + 64x5 + (16c + 16)x4 + (8c + 16d)x3 + • • • .
Then equating the corresponding coefficients of x4 and of x3, we get 16c 4-16 = -48 and 8c + 16d = -48, respectively.
Solving for c and d, we find c = -4 and d - -1. Hence,
64x6 + 64x5 - 48x4 - 48x3 + 8x2 + 8x + 1 = (8x3 + 4x2 - 4x - l)2,
which we can verify by expanding.
Therefore, a cubic whose roots are cos y, cos y, and cos y is 8x3 4- 4x2 - 4x - 1
7.73 We see that z,z2,, z2005 are the 2006th roots of unity other than 1.
Let co be a 2006th root of unity other than 1. Then co2006 - 1 = 0, which factors as
(co - l)(a;2005 + co2004 + co2003 + • • • + CO + 1) = 0.
Since co =£ 1, co satisfies the equation a;2005 + co2004 4- co2003 4-+ co + 1 = 0.
We have
f(co) = co2004 4- 2a;2003 + 3co2002 + • • • + 2004a; 4- 2005.
This is similar to a geometric series, so we try using the same strategy to evaluate it. We multiply this equation by
co to get
cof(co) = CO2005 + 2co2004 + 3co2003 + • • • + 2004a;2 + 2005a;.
Subtracting the equation for f(co) from this equation gives
(co - l)f(co) = co2005 4- 2a;2004 + 3a;2003 4- • • • + 2004a;2 + 2005a; - co2004 - 2co2003 - 3a;2002 -2004a; - 2005
= co2005 + co2004 + co2003 4- • • • 4- o - 2005
= (co2005 + CO2004 + CO2003 + ■ • • + Q) 4-1) - 2006
= -2006,
so f(co) = = yz*2. Since this holds for any 2006th root of unity other than 1, we have
N = mfh?) ■ ■ ■ flz2005) = . 20“ ... _2006^_

" ' ■' ’ 1 — Z l-z2 1 -z2(j® (1 - 2)(1 - Z2) - • • (1 - 22005)'
Since 1, z, z2,..., z2005 are the 2006th roots of unity, we have
(x - l)(x - z)(x - z2) • • • (x - z2005) = x2006 - 1.
But x2006 - 1 factors as (x - l)(x2005 4- x2004 4- ••• 4- x 4-1), so
(x - z)(x - z2) • • • (x - z2005) = x2005 4- x2004 4- • • • 4- X 4- 1.
148
Challenge Problems
Taking x = 1, we get (1 - z)(l - z2) • • • (1 - z2005) = 2006. Therefore,
20062005 20062005 2004

N= 2006
(1 - z)(l - z2) • • • (1 - z2005) " 2006
7.74 The roots of z1997 1 — 0 are the 1997th roots of unity, which are of the form g2*71'/1997, where 0 < k < 1996.
Suppose v = e2)71'/1997 and w = e2kml1997. We then have
\V + W\= |e2ira'/l"7 + ^kni/19971 _ ^Zjni/1997^ ^ + e2(k-j)ni/\997\ _ ^jni/1997^ + ^2(k-j)ni/19971 _ ^ + e2(k-j)ni/19971
Since c2^ /W1"7 ig also a solution to z1997 — 1 = 0, the probability that \v + iv\ > \]l. + V3 equals the probability
that |1 + f| > \/2 + a/3, where t is also a nonreal solution to z1997 — 1 = 0. (We exclude t = 1 because zz and w must
be distinct, so k t j.)
We let t = 1 )97 for some a, where 1 < a < 1996. (The argument of t cannot be 0, because v and w are distinct.
Thus, there are 1996 possible values of a.) Let a = Then t = e2ani/'1997 - eia, and
|1 + f| = |1 + em\ = |1 + cos a + i sin a\ = \/(l + cos a)1 + sin2 a = \/1 + 2 cos a + cos2 a + sin2 a = V2 + 2 cos a.
Hence,
V3
|1 + f| > ^2+ V3 V2 + 2 cos a > \/2 + V3 <=> 2 + 2 cos n > 2 + V3 <=> cos a >
Since 0 < a < 2n, the inequality cos a> ^ holds if and only if0<n<|ori|ZI<a< 2n.
Since a =
_ n 2na n 1997
0<a<- <=> 0 < —— < — o 0<a<-.
6 1997 6 12
Since a is an integer, we have 1 < a < 166, for 166 possible values of a. Similarly,
lln „ ll7i 2na n 21967

— <a<2n « —< — <271 « — < a < 1997.
Since a is an integer, we have 1831 < a < 1996, for another 166 possible values of a.
Therefore, the probability that \v + w\> \]2 + V3 is (166 + 166)/1996 = 332/1996 = 83/499
7.75 We turn this into a problem involving complex numbers by recalling that sin 9 = , so
ni/n _ p-m/n \ 7 p2ni In _ p-2ni/n \ / p3ni/n _ p-3ni/n ,(n-l)ni/n _ g-(n-l)ni/n

.77.271. (n - 1)77
sm — sm-sin-
n n n 2i 2i 2i 2i
That's still pretty scary. But, we can do a little factoring to make the numerators expressions that we're more
comfortable with:
,ni/n _ g-ni/n \ / e2ni/n _ e~2ni/n \ 7 £3ni/n _ e~3ni/n y(n-l)ni/n _ g-(n-l)ni/n
2i 2i J V 2z y V 2T"
e-ni/nse2ni/n — 1)\ / g-2ni/nîni/n — 1)\ /g-3ra7”(g6m'/” - 1) e~(n-l)ni/n^^2(n-l)ni/n _
2z 2/ 2/ (♦)
2z
149
Aha! The factors in the numerator look suspiciously like factors of x" - 1. Let's investigate. Let co = e2m/n. The
roots of xn - 1 = 0 are the nth roots of unity, namely 1, co, co2,..., con~l. The polynomial xn - 1 factors as
xn - 1 = (x - IK*""1 + xn~2 + • • • + x + 1).
Hence, the roots of x”-1 + x"~2 H-1- x + 1 = 0 are <n, co2,..., a;"-1, which means
(x - co){x - co2) ■ ■ ■ (x - co"-1) = xn_1 + xM_2 + • ■ • + X + 1.
Taking x = 1, we get
{1 - co)(l - co2) ■ ■ ■ (1 - co"'1) = n.
We tie this to the product of sines by reversing the factoring we did on line (*). We do so by noting that
1 — eie = g*0/2(g-*0/2 _ g'0/2) - _g!0/2^gZ0/2 _ g-z'6/2^ so
,-z0/21
|1 - eie| = \-ei9l\ei9'2 - e-i9/2)\ = \-eie'2\ \eie/2 - e~i9/2\ = |e;e/2 - e"1' 1
Since sin 0 = e'e 2- 's, we have

\eW/2 _e-id/2\ _ 0
2z sm — = 2 sm
2
Since 0 < f < 7i, we have sin | > 0, so |1 — el6\ = sin f. In particular, for 0 = where 1 < k < n - 1, we have
cok - e2'kniln = el9, so
kn
|1 - cok | = 2 sin
n
Therefore, taking the magnitude of both sides of the equation (1 - &;)(! - co2) con x) = n, we get
r. . n 2n (n-l)n
2 sm — • 2 sm — ■ • • 2 sm-= n.
n n n
n 2n (n-l)n _ n
Finally, dividing both sides by 2n 1, we get sin ~ sin ^ • • • sin [n n1)n - as desired.
150
CHAPTER
I -1
Geometry of Complex Numbers
8.1.1 Since w is a counterclockwise rotation of z by n, we have w = em(z - 0) + 0 = -z . Note that this means
that w + z = 0, so the origin is the midpoint of the segment connecting w and z. We therefore say that u> is the
"reflection of z through the origin."
8.1.2
(a) The image of 3 + z upon a 30° (or |) counterclockwise rotation about the origin is
,0 3 V3 V3. 3. 1 3V3-1 3 + V3.

em/6(3 + i) = (3 +1) = + ~l +-!-- = --- H---1
2 2 2 2
(b) The image of 3 + i upon a 120° (or y) clockwise rotation about 1 + 2i is
V3 5-2V3.
e-2ni/3[(3+i)-(1+2/)]+(1+2z) = (2-i)+l+2/ = -l + ^-V3i-^+l+2i =
“2“ +-2-Z
8.1.3 It is not true that the argument of w + z is always the average of the argument of w and the argument of z.
For example, take w - 2 and z = i. The argument of w - 2 is 0, and the argument of z = i is but the argument of
w + z = 2 + i is clearly not |.
However, it is true that if \w\ = |z|, then the argument of w + z is the average of the Im
argument of w and the argument of z. We can see this geometrically.
w+z
The origin, w, w + z, and z always form a parallelogram. But when \w\ = |z|, this
parallelogram is a rhombus. Hence, the line joining the origin and w + z bisects the
angle of the parallelogram at the origin. In other words, the argument of w + z is the
average of the argument of w and the argument of z.
151
CHAPTER 8. GEOMETRY OF COMPLEX NUMBERS
We can also use algebra. We let w = rela and z = re'P. We then have
w + z = r(eia + e'P)
= r (cos a + cos j3 + z(sin a + sin jS))
a- j3 a + (3 a - j8
= 2r (^cos cos +z cos
2 2 2
a - /3 a+ n + jS
= 2rcos cos + zsm
V 2
a ^ J(a+p)/2
= 2 r cos
(Note that we used the sum-to-product identities for sine and cosine as an intermediate step.) Since 2r cos(^y) is
real, the argument of w + z is ^y.
8.1.4 Since
-id
= -e
reid
the transformation that maps z to z/(rel9) is a combination of dilating about the origin with a scale factor of 1/r
and rotating by an angle of 9 clockwise about the origin .
8.1.5
(a) The function / corresponds to a translation of -5 units horizontally and 2 units vertically.
(b) The function / corresponds to a dilation about the origin with scale factor 3, followed by a translation of
-1 unit horizontally. (Note that f(z) = 3z - | = 3 (z - | j, so we can also describe the transformation as
a dilation about ~ with scale factor 3.)
(c) Since \ + -y^z = the function / corresponds to a rotation about the origin counterclockwise by an angle
off.
(d) Since -1 + z = V2e3ra/4, the function / corresponds to a dilation about the origin with a scale factor of V2,
followed by a rotation about the origin counterclockwise by an angle of ^ (or vice versa).
(e) Since -z = e3ra/2, the function / corresponds to a rotation about the origin counterclockwise by an angle of
followed by a translation of -1 unit horizontally and 3 units vertically.
We can also describe / as a rotation about a point. We see this by noting that if we rotate z about w by
an angle of 9 counterclockwise, then the image is el6(z — w) + w. Comparing this to f(z) - -iz - 1 + 3z, we
see that we can find values of w and 9 such that el0(z — w) + w = -iz - 1 + 3z. Expanding the product on the
left, we have
el9z - el9zo + w = -iz - 1 + 3z.
Since this must be true for all z, we must have et9 = -z, so we can choose 9 = y, and our equation is
-iz + iw + w = -iz - 1 + 3z.
Solving for w gives
-1 + 3z _ (-1 + 3z’)(l - z) —1 + 3z + z + 3 2 + 4z
= 1 + 2 z.
1 +z (1 + 0(1 — z) 2 2
Hence, / also corresponds to a rotation about 1 + 2z counterclockwise by an angle of y.
152
Section 8.2
8.1.6 We translate the points W, U, and Z to W, If', and Z', so that

W coincides with the origin. Then W, IF, and Z' are collinear, and
U'W'/U'Z' = a/b. Then U'W'/W'Z' = a/(a+b). The corresponding complex
numbers are zv' = zv - zv = 0, u' = u - zv, and z' =z-w.
We can think of u' as the result of a dilating z' through the origin with
scale factor a/(a + b), so
Re
Substituting u' = u-w and z' = z - zv, we get
a
u-zv - (z - w),
a+b
so
az-aw + aw + bio az + bw
u =--(z - zv) + zv =
a+b a+b a+b
8.1.7 We have z2 = em/2Zi - iz\, and
z3 = em/2(z2 - 4 - 3i) + (4 + 3i) = i(izi - 4 - 3z) + (4 + 3i) - -z\ - 4i + 3 + 4 + 3i - -z\+7 - i.
If Z3 is the point obtained when z\ is rotated counterclockwise about zv by an angle 8, then
je. - eww + zv.

Z3 = e'e(z\ -w) + w = ewz\
Since Z3 = — Z\ +7 - i from above, we seek 8 and zv such that el9Z\ - el6w + 10 = —Z\ + 7 -i. This means we must
have el9 - -1 and -e,ew + zv = 7 -i. We can take 6 = [~tT\ so -el9w + w- 2zv, which gives w = ^ = z2 - if
2l
.
8 2.1 The points in the complex plane that represent 1 + 3z, 5 - 2i, and z are collinear if and only if is reah
so we check this value for each the given complex numbers.
For z = 3 - 2i, we have = ^=f1 = ~1 + \h which is not real.
For z = 9 - 7i, we have = 2, which is real.
(—2+5Q—(1+3/) _ -3+2/ _ -3+2/ (—3+2/)(—1—/)

For z = -2 + 5i, we have (-2+5/)—(5—2/) -7+7/ 7(-l+/)
= which is not real.
7(—!+/’)(—1—*)
Therefore, of the given complex numbers, only 9 - 7/ is collinear with 1 + 3z and 5 - 2i.
.
8 2.2 We showed in the text that complex numbers z, a, and b are collinear if and only if is real. But a complex
number is real if and only if it is equal to its own conjugate. The conjugate of is Hence, z, a, and b are
collinear if and only if
z — a _ z. - a
b-a b-a‘
153
8.2.3 Let the vertices of the quadrilateral be a, b, c, and d. Then the midpoints B
of the sides are pr, and The quadrilateral with these points as
vertices is dashed at right. In this quadrilateral, the midpoint of the diagonal
joining ^ and ~ is
a+b . c+d
2 'r 2 a+b+c+d
and the midpoint of the diagonal joining b+c

2 and is
a+b+c+d
Thus, the midpoints of the two diagonals coincide. In other words, the two diagonals bisect each other. Therefore,
the quadrilateral formed by the midpoints of the sides is a parallelogram.
8.2.4 If OB' ± OD', then we can rotate D' by 90° to produce a point that is on the ray from the origin through
B'. (That is, the complex number corresponding to the image of the rotation has the same argument as b'.) The
complex number corresponding to the image of this rotation then is either id' or -id', depending on the orientation
of the 90° rotation. Since the image is on the ray from the origin through B', there is a dilation that maps this point
to B'. Therefore, we have b' = k(id') or b' = k(-id') for some real constant k. In either case, b'/d' is imaginary.
8.2.5
(a) The complex number corresponding to the midpoint of BC is (b + c)/2. So, to show that the point corre¬
sponding to (a + b + c)/3 is on the median from A to BC, we must show that a, (a + b + c)/3, and (b + c)/2 are
collinear. We have
a+b+c
-a _ b-^ _ 2
b+c b+c-2a
-a 3'
2 2
which is real. Therefore, a, (a + b + c)j3, and (b + c)/2 are collinear. (Note that we cannot have b + c - 2a = 0
because this gives us a = (b + c)/2, which would mean that A is the midpoint of BC. Since ABC is a triangle,
A cannot be the midpoint of BC.)
Similarly, we can show that G is on each of the other two medians, so the medians of AABC meet at
the point corresponding to (a + b + c)/3. (This matches our intuition: the complex number corresponding
to the midpoint of a segment is the arithmetic mean of the numbers corresponding to its endpoints, and
the complex number corresponding to the centroid of a triangle is the arithmetic mean of the numbers
corresponding to the triangle's vertices.)
(b) Our work in the first part pretty much covers this part, too. Letting M be the midpoint of BC, we have
a+b+c a b+c-2a 1
AG _ ; | _ | 3 1 2 \b + c-2a\ _ 2
AM " \*f-a\ b+c-2a 1 3 ’ | b + c- 2a\ ~ 3'
2 1
so AG is | of AM, as desired. There's nothing special about the median from A, so the distances from B and
C to the centroid are also equal to | of the lengths of the respective medians from B and C.
8.2.6 As shown in the text, a, b, and c are collinear if and only if (a - b)(b -c) - (a - b)(b - c) = 0. Expanding the
products on the left gives
ab - ac - bb + be -db + ac + bb - be - 0.
Cancelling the bb terms and rearranging gives the desired
ab + be + ca - db + bc + ca.
8.2.7 Let w, x, y, and z be complex numbers corresponding to W, X, Y, and Z. Then the midpoints of WX, WY,
and XZ are and In the text, we showed that complex numbers a, b, and c are collinear in the complex
154
Section 8.2
plane if and only if

(b - a)(b - c) - (b - a)(b - c) = 0.
We take (x+w)/2to be b, since it has terms in common with both other points, and we'll therefore have cancellation
in each term in parentheses above. Because the three given midpoints are collinear, we have
/iv + x w + y\ /w + x x + z\ (w + x w + y\ fiv + x x + z^_n

' 2 2 / V 2 ~2~ J~\ 2 2 ) V~2 2~ / ~ '
Simplifying each of the differences gives us
- ({x - y)(w -z)-(x- y){w - z)) = 0, (8.1)
so (x - y)(w -z)-(x- y)(w - z) = 0.
The midpoint of YZ is To check if it is on the line through the other three midpoints, we use our test
for collinearity on b = , a - and c = Tp. (Notice that we choose our points strategically to produce
cancellation in the test for collinearity.) We have
,, w-r —\ ,r _w, N (V + z w + y\ fy + z YTz\ /y + z w + y\ (y + z x + z\

((, _ a)(b-c)-(fc-«)(b -c) = -2 ) V 2 2~/ _ \2-2 ) ( 2 ~ ~)
1 _____
= 4 ((z- w)(y -x)-(z- w){y - x))
= \ ((w " z)(x -y) - (w - z)(x - y))
= 0,
where we apply the result of Equation (8.1) to provide the last step. Therefore, the midpoint of YZ is indeed on
the line through the other three midpoints.
8.2.8 Let a, b, c, and d be the complex numbers corresponding to A, B, C, and Im

D, respectively. First, we translate A and B to A' and B', so that A' coincides
with the origin O. The corresponding complex numbers are a' = a - a = 0
and b' = b - a = (3 + 2i) - (1 - 5z) = 2 + 71
Next, we translate C and D to C' and D', so that D' coincides with the
origin O. The corresponding complex numbers are d' = d - d = 0 and
d =c-d = (-7 V3 + 7i) - (-2 - 2i V3) = (2 - 7 V3) + (7 + 2 y/3)i. Note that the Re
angle between AB and CD is equal to IB'OC'.
Multiplying a complex number by re'6 corresponds to the combination of a

dilation (about the origin) by a factor of r and a rotation of d counterclockwise
about the origin. Hence, iB'OC is the argument of c'/b'.
We have
d (2-7V3) + (7 + 2V3)z _ [(2 - 7 a/3) + (7 + 2 V3)z](2 - 7i)

b'~ 2 + 7i (2 + 7z)(2 - 7i)
(4 - 14 V5) + (-14 + 49 V3)z + (14 + 4 V3)z + (49 + 14 V3) _ 53 + 53 V3t
= ^ 53 _ 53
In exponential form, we have 1 + V3z = 2 + ^z) = 2eni/3, so the acute angle between and ® is
155
8.3.1 Any circle in the complex plane is the graph of an equation of the form |z — c\ = r, where c is a complex
number and r is a positive constant. As described in the text, we can square both sides and write this equation in
the form (z - c)(z-c) = r2, or
zz — cz - cz + cc = r2.
Comparing this to the given equation, we have -c = -2 - i from the coefficient of z, so c = 2 + i. Then, we have
cc = 5, so the equation for the circle above is zz - (2 - i)z - (2 + i)z + 5 = r2, or
zz + (-2 + i)z + (-2 - i)z = rz -5.
Comparing this to the given equation, we have r2 - 5 - 11, so r2 - 16, which gives r = [4].
8.3.2 Let a, b, c, and p be complex numbers corresponding to points A, B, C, and P. Then, we wish to show that
|a - c\2 — \c — p|2 = \a - b\2 — |b — p\2, which we can also write as
|a - c|2 - \c - p|2 - |a — b\2 + \b - p\2 = 0.
We can make life easier on ourselves by choosing one of these points to be the origin. We'll choose p to be the
origin, so we wish to show that
|a - c|2 - |c|2 - |a - b\2 + \b\2 - 0.
Since \z\2 = zz for any complex number, we have
|a - c\2 - \c\2 - |a - b\2 + \b\2 = (a - c)(a - c) - cc - (a - b)(a -b) + bb
= {a- c)(a - c) - cc — (a- b)(a -b) + bb
= aa - ac - ca + cc - cc - aa + ab + ba -bb + bb
- —ac - ca + ab + ba
= a(b - c) + a(b - c).
Because^ J_ ^,wehave(a-p)(b-c)+(a-p)(b~c) = 0. Because we letpbe the origin, we havea(b-c)+a(b-c) =

0, so our expression for |a - c|2 - |c|2 — |a - b\2 + \b\2 equals 0, as desired. Therefore, AC2 - CP2 = AB2 - BP2.
8.3.3
(a) By the Triangle Inequality, we have \x + y - z\ + \x - y + z\ > |(x + y - z) + (x - y + z)| = |2x| = 2\x\.
(b) Similarly, \x + y - z| + \-x + y + z\ > \{x + y -z) + (-x + y - z) | = |2y| -- 2|y|, and \x - y + z\ + \-x + y + z\ >
|(x - y + z) + (—x + y + z)\ = |2z| = 2|z|. Adding these two and the result from part (a), we get
2\x + y - z\ + 2\x - y + z\ + 2\-x + y + z\ > 2\x\ + 2|y| + 2|z|.
Dividing both sides by 2 yields |x + y - z\ + |x - y + z\ + |-x + y + z| > |x| + |y| + |z|.
8.3.4 Because \w\ = 1, the point corresponding to w in the complex plane is on the unit circle. Similarly, w + a and
w + bi are on the unit circle. Let the points corresponding to w, w + a, and w + bi be W, A, and B. Since w + a is a
horizontal translation of w, point A is directly to the left or right of W. Similarly, since iv + bi is a vertical translation
of W, point B is directly above or below W. Therefore, ZAWB is a right angle. Because lAWB is a right angle that
is inscribed in a circle, the arc of this circle connecting A to B is a semicircle. Therefore, AB is a diameter of the unit
circle, which means its length is 2. From the Pythagorean Theorem, we also have AB2 = WA2 + WB2 - a2 + b2, so
a2 + b2 = AB2 = 4.
8.3.5 Let the quadrilateral be ABCD and let a, b, c, and d correspond to A, B, C, and D. Then, we wish to show
that
\a - c\ +\b - d\ <\a - b\ + \b-c\ + \c - d\ + \d- a\.
156
Section 8.3
We see two different pairs on the right side to which we can apply the Triangle Inequality in order to produce an
I a - c\ term on the lesser side. We have
\a-b\ + \b-c\> | (a -b) + (b - c)| = | a- c\,

\c - d\ + \d - a\> |(c - d) + (d - a)\ = \c - a\ - \a - c\,
both by the Triangle Inequality. Note that we can write both inequalities as strict inequalities because ABC and
ACD are triangles, so A cannot be on BC or CD. Adding these two inequalities gives us
\a-b\ + \b - c\ + \c - d\ + \d - a\> 2\a - c\. (*)
We can do the same thing with |b — d|. The Triangle Inequality gives
\b-c\ + \c-d\> \{b -c) + (c- d)| = |b- d\,

\d-a\ + \a-b\> \{d -a) + (a- b)\ = \d-b\ = \b- d\,
where again we have strict inequality because we cannot have any three vertices of the quadrilateral be collinear.
Adding these inequalities gives
\a-b\ + \b - c\ + \c - d\+ \d - a\> 2\b - d\,
and adding this inequality to inequality (A), and then dividing the resulting inequality by 2, gives the desired
\a - b\ + \b - c\ + \c - d\ + \d - a\ > \a - c\ + \b - d\.
8.3.6 Since both |w + z| and \w\ + |z| are positive, we can compare |w + z|2 and (|w| + |z|)2 in order to compare |w + z|
and \w\ + |z|. We introduce the squares because they are easier to work with. We have
(|w| + |z|)2 = \zv\2 + 2\w\\z\ + |z|2 = ww + 2|ie||z| + zz,

|iv + z|2 = (w + z)(w + z)- ww + WZ + ZW + ziu.
Therefore, we have
(|w| + |z|)2 - |w + z|2 = 2|ze||z| - (wz + zw).
We'd like to simplify wz + zw. Where have we seen an expression like this before? We first note that wz and zw
are conjugates:
wz-w-z - wz.
Then, we recall that the sum of a complex number and its conjugate is twice its real part: a + bi + a + bi = 2a. So,
we have
(|ie| + |z|)2 - \w + z|2 = 2\w\\z\ - 2Re(wz).
We want to show that the expression on the right is nonnegative, so we wish to show that |ze||z| > Re(zez). So that
we'll be working with like expressions, we note that |z| = |z|, so |ze||z| = |ze||z| = \wz\, which means we'd now like
to prove that \wz\ > Re(wz). Now that we're comparing apples to apples, the result is obvious. The magnitude of
any complex number is at least as large as its real part:
|a + bi\ = Vfl2 + b2 > Vfli2 = \a\ > a.
Note that equality holds here if and only if b = 0 and \a\ = a; that is, if the complex number is a positive real
number.
We therefore have \wz\ > Re(zez), which means \w\\z\ > Re(wz), so
(|ze| + |z|)2 - |w + z|2 = 2|w||z| - 2Re(wz) > 0,
157
and we can conclude that \w\ + |z| > \w + z\.
Equality holds in \w\ + |z| > \iu + z| if and only if it also holds in our key inequality step, \wz\ > Re(zvz). As
explained above, this inequality only holds if wz is a positive real number. Multiplying wz = c by z gives wzz = cz,
so f ~ W' ^^rice c *s positive, this tells us that w and z are on the same ray from the origin.
Putting this all together, we have \w\ + |z| > \zv + z|, where equality holds if an only if w/z is a nonnegative real
number or z = 0.
8.4.1 Let r\ and r-i be the complex numbers corresponding to the two possible locations of R,
which are points Pi and R? shown at right. Ri
Then Ri is obtained by rotating P clockwise around Q by so
r\ - e~m/2(p -q) + q = -i{p - q) + q - -ip 4- (1 + i)q
Also, R2 is obtained by rotating P counterclockwise around Q by |, so
r2 = em/2(p -q) + q = i(p-q) + q = ip + (1 - i)q
8.4.2 Let a be the complex number corresponding to A, and similarly for the other
points. Let C be a primitive sixth root of unity, so C2 = C — 1, and let O be the origin.
Then b = C,a,d = C,c, and / = L,er so
b+c Ca + c d+e Cc + e f+a C,e + a

l m --—, n = = —-—.
To show that triangle LMN is equilateral, it suffices to show that n-l = C(m - /).
Substituting, we find
C,e + a C,a + c (1-Qa-c + Cfi

n-l =
2 2 2
and
C,c + 6 C,a + c -Cfl + (—1 + C^c + 6
m—l =
2 2
so
-C 2a + (-C + C2)c + Ce (1 -Qa-c + &
C(m -1) =
where we use the fact that C2 = C — 1 in the final step. Hence, n-l - £(m - l), so triangle LMN is equilateral.
8.4.3 Let a be the complex number corresponding to A, and similarly for the other points.
We let b be the origin, so that d = a + c. Next, we find / and g in terms of a and c.
Because AABG is equilateral and B is the origin, we have g - La for some sixth root of
unity £• From equilateral triangle CFB, we have c - £/. We are careful to choose the same
orientation of the rotation as we chose when considering A ABG, so that the sixth root of
unity in g = Lais the same as the sixth root of unity in c = £/. Solving for / gives / = c/C-
158
Section 8.4
Now, in order to check if A DGF is equilateral, we must check if d — g = C,(/ — g), where we again strategically
choose to check the rotation of F counterclockwise about G because doing so will allow us to cancel the C hi the
denominator of / = c/G
We have
d - g = (a + c) - Cfl = (1 - Qa + c,
£(/-#) = C (| - 0*) = c - C2a-
As described in the text, we have £2 - C + 1 = 0, so C2 = C - 1, and we have
C(/ - g) = c - C2fl = C - (C - l)fl = (1 - Qa + c = d - g,
which matches our expression for d - g from above. So, triangle DFG is equilateral.
8.4.4
(a) The relationship we must prove looks a lot like the condition we proved in the text under which a triangle
is equilateral. There, we showed that A ABC is equilateral if and only if a — b = Q(c - b) for some primitive
sixth root of unity G We make this look even more like a-b - co(b - c) by writing it as a — b - —C,(b - c).
Now, we just have to show that —Q is a primitive cube root of unity if and only if C is a primitive sixth
root of unity. If C = e71^3, we have
-C = (~l)eni/3 = eni • eni/3 = eAni/3,
which is a primitive cube root of unity. Similarly, if C = e5n'13, then —C = e2ni13, which is also a primitive
cube root of unity. We conclude that —C is a primitive cube root of unity if and only if C is a primitive sixth
root of unity. Combining this with the fact that a - b = ~C (b - c) if and only if A ABC is equilateral, we see
that A ABC is equilateral if and only if a-b = co(b - c) for some primitive cube root of unity co.
The diagram at right gives us some geometric intuition for the relationship Im
between the primitive cube roots and the primitive sixth roots of unity. The
vertices of regular hexagon ABCDEF are the sixth roots of unity. Points B and
F correspond to the primitive sixth roots. Points C and E correspond to sixth
roots of unity that are also primitive cube roots of unity. Each primitive sixth
root of unity is the image of rotating a primitive cube root of unity by 180° about
the origin. Therefore, each primitive sixth root of unity is the negative of some
primitive cube root of unity.
(b) We'd like to use the previous part to knock off this part, since part (a) already has a primitive cube root of
unity in it. However, we need a way to introduce co2. We could multiply both sides of a-b - co(b - c) by co,
but that gives us co(a - b) = co2(b - c), which doesn't appear to be much help. If we rearrange this equation,
we can group the b terms to give
coa - (co + co2)b + co2c - 0.
Where have we seen co + co2 before when thinking about cube roots of unity? In the text, we showed that
the primitive cube roots of unity are the roots of z2 + z + 1, which means that co2 + co + 1 = 0, so co2 + co - -1.
Therefore, our equation coa - (co + co2)b + co2c = 0 becomes coa + b + co2c = 0. So, A ABC is equilateral if and
only if b + coa + co2c = 0. Uh-oh. That's not quite what we wanted. We wanted a + cob + co2c = 0. Is that a
problem?
No, it's not. We just showed that if (a - b) = co{b - c), then b + coa + co2c. From part (a), we also know
that we must have (b - a) = co'(a - c), where all we have done is swapped a and b, for some primitive
cube root of unity co'. If we follow the same steps we used above to show b + coa + co2c - 0, but start with
(b - a) = co'(a - c), we will end at a + co'b + (co')2c = 0. Since co' is a cube root of unity, we have proved the
159
desired relationship. (What's really going on here is that co and co' are not the same primitive cube root of
unity, which is why starting with (a-b) = co(b - c) doesn't get you exactly what you want, but starting with
(b - a) = co'(a - c) does.)
8.4.5 Let a be the complex number corresponding to A, and simi¬ B

larly for the other points. We place the square ABCD in the complex
plane so that a = 1, b = z, c = -1, and d = -z. Then the center of the
square O coincides with the origin.
Then K is obtained by rotating B by | counterclockwise around

A, so
?ni/3(b -a) + a=(\ + (i -1) + 1 = 1—'— + 1—^z.
By symmetry, L can be obtained by rotating Kby f counterclock¬

wise around the origin, so
A - V3 1— V3.\ V3-1 1-V3.

\ 2 + 27“ 2 + 2*
Since M and N are rotations by tl of K and L about the origin, we

have m = -k = and n = -l =
Let Pi be the midpoint of ML, so pi - ^ . Let P2 be the midpoint of AN, so
a+n 1 + 1-V3 + 2 3-V3 V3-1

V2
2
Then
P2 = + = (3- y/3) + (V3-l)i = V3(V3-l) + (V3-l)z = V3 + f
Pi 2(V3-1) 2( a/3 — 1) 2 '
which is equal to
V3 1 . Tl . . 71 m76
— + - z = cos — + z sm — = e 1 .
2 2 6 6
Hence, P2 can be obtained by rotating Pi by | counterclockwise about the origin.
Now, let P3 be the midpoint of BL, and let P4 be the midpoint of MN. Then by symmetry, P4 can be obtained
by rotating P3 by | counterclockwise about the origin. Furthermore, OPi = OP4, and ZP1OP4 = |, so OP2 = OP3
and /P2OP3 = 2_6~6 = 6- ênce/ ^1/ ^2/ P3/ arid P4 are four of the vertices of a regular dodecagon, centered at
O.
By symmetry, all 12 midpoints form the vertices of a regular dodecagon.
Review Problems
8.32 If z = x + yi, then z = x - yi. Thus, z has the same real part as z, but the imaginary parts of z and z are
opposites. Hence, reflection over the real axis maps z to z.
8.33 We see that f\(z) = z + 2 - 3z and fj(z) = em/2(z - 4z) + 4z = (z)(z - 4z) + 4z = zz + 4 + 4z
160
Review Problems
(a) We have (/2 o /x)(z) = /2(/i(z)) = /2(z + 2 - 3/) = i(z + 2 - 3/) + 4 + 4z = zz + 7 + 6 z
(b) We have (/i o /2)(z) = /i(/2(z)) = /j(zz + 4 + 4z) = zz + 4 + 4z + 2 - 3z = zz + 6 +z
Note that our results for parts (a) and (b) are not the same! We did not make a mistake; this is an example of the
fact that a translation followed by a rotation is not necessarily the same as doing the rotation first, and following
it with the translation. (See if you can prove that the only time these two are the same is when one of the two
transformations does nothing, such as a rotation of 2n. Here's a hint: think about what happens to the origin.)
8.34 Let z = x + yi. Then f{z) - zz = z(x - yi) = y + xi. Thus, / swaps the real and imaginary parts of z. The
transformation corresponding to / is reflection over the line x-y
8.35 In both parts, let points A, B, and C correspond to a, b, and c, respectively. Then, we have \a - b\ = AB,
\a-c\= AC, and \b - c\ = BC.
(a) We have AB + AC - BC. By the Triangle Inequality, this equation holds if and only if A is on BC.
(b) We have AB2 + AC2 = BC2. By the Pythagorean Theorem, this equation holds if and only if ABAC is a right
angle.
8.36
Im
(a) First, note that the magnitude of ^ is = |[ = 1, so fj lies on the unit circle.
|z|
Furthermore, the complex number |j can be obtained by dilating z through the
origin by a factor of pj. Hence, p is the intersection of the unit circle and the ray
joining the origin to z.
(b) Let z = x + yi, where x and y are real numbers. Then z = x - yz, so z + z = lx.
Hence, the foot of the altitude from z to the x-axis, which we'll call point M,
is the midpoint of the segment from the origin to z + z. After locating M, we
draw the circle with center M and radius OM, where O is the origin. The other
point at which this circle intersects the real axis corresponds to z + z. Re
By definition, z lies on the circle centered at the origin with radius |z|. Im
(c)
JL
Furthermore, |z| is a real number, i.e. |z| is the intersection of this circle
and the nonnegative real axis. The point corresponding to |z| is |z| to the
right of the origin. The point corresponding to z + |z| is |z| to the right of
z. Therefore, if we build a quadrilateral with vertices 0, z, z + |z|, and |z|
in the complex plane, the side connecting 0 to |z| and the side connecting
z to z + |z| are parallel and equal in length. Therefore, to locate z + |z|, we
draw the line through |z| that is parallel to the line from the origin to z. The
intersection of this line with the line through z parallel to the real axis is the
point corresponding to z + |z|. W
161
8.37 Let A, B, and C be the vertices of the triangle, and let a, b, and c be the corresponding A
complex numbers. Let M and N be the midpoints of AB and AC, respectively, so the
corresponding complex numbers are m - ^ and n = respectively.
We therefore have m - n - ^. This means that MN - \m - n\ = \^y\ =

^\b - c\ = ^p, and that Since is real we know that MN || BC. Therefore, the
segment connecting the midpoints of AB and AC is parallel to BC and has half the length of
BC.
8.38
(a) Let a = 3 - 2z and b - 1 - 5z. As shown in the text, the line joining a and b is perpendicular to the line joining
z and b if and only if
z-b _ z - (1 - 5i) _ z - (1 - 5i)
a^b ~ (3 - 2z) - (1 - 5f) _ 2 + 3z
is imaginary, so we check this value for each the given complex numbers.
For z = 2 — 7z, we have
(2 - 7i) - (1 - 5f) _ 1 — 2i _ (1 — 2i)(2 - 3z) _ 2 - 3i - 4z - 6 _ -4 - 7/

2 + 3z ~ 2 + 3z " (2 + 3z‘)(2 - 3z) " 13 ~ 13 '
which is not imaginary.

For z = -6 + 2z, we have
(-6 + 2z') - (1 - 5z') _ -7 + 7z _ (-7 + 7z')(2 - 3z‘) _ -14 + 21z + 14/ + 21 _ 7 + 35/
2 + 3/ ~ 2 + 3/ “ (2 + 3z‘)(2 - 3i) ~ 13 ~ 13 '
which is not imaginary.

For z — —5 — z, we have
(-5 - 0 - (1 - 5z‘) _ -6 + 4/ _ (-6 + 4z')(2 - 3z) _ -12 + 18/ + 8/ + 12 _ 26/ _ „.

2 + 3/ ~ 2 + 3/ ~ (2 + 3z)(2 - 3i) ~ 13 “ 13 “ h
which is imaginary.
Therefore, of the given complex numbers, only | -5 — z has the desired property.
(b) If z is on the line through b such that the line is perpendicular to the line through a and b, then we must
have
(z - b)(a -b) + (z- b)(a - b) = 0,
as proved in the text. Therefore, the desired equation is
(z - (1 - 50)((3 + 2z') - (1 + 50) + (z - (1 + 5z'))((3 - 2i) - (1 - 50) = 0.
Simplifying the left side gives
(z - 1 + 5z‘)(2 - 3z') + (z - 1 - 5z')(2 + 3z) = 0.
Expanding and simplifying the left side gives (2 - 3z')z + (2 + 3i)z + 26 = 0
8.39 Let a, b, and d be complex numbers corresponding to points A, B, and D in the complex plane. We know
that any circle in the complex plane can be described as the graph of some equation |z - c\ = r, where c is the center
and r is the radius. We can make our work a lot easier if we choose the midpoint of AB to be the origin, so the
circle through A and B in the complex plane has center 0 and radius \a\. Therefore, the circle with diameter AB is
the graph of the equation |z| = \a\. So, to show that D is on the circle, we must prove that \d\ = |a|.
162
Review Problems
If AD ± DB, then we have

(a - d)(b -d) + (d- d)(b -d) = 0.
Since we chose the midpoint of AB to be the origin, we have b - -a, and the equation above becomes
(a - d)(-a -d) + (a- d)(-a - d) - 0.
Expanding the left side, we have
-ad - ad + da + dd - ad - dd + ad + dd - 0.
Massive cancellation ensues, and we're left with
2dd - 2aa = 0.
Dividing by 2 and rearranging gives dd = ad, from which we conclude that \d\2 = |a|2 (because zz - \z\2 for any
|z|). Finally, since \d\ and |a| are nonnegative, we can take the square root to find \d\ - \a\, as desired. Therefore, we
conclude that D is on the circle with diameter AB.
8.40 The center of the hexagon is given by (2+;)+2(8 3?) _ 5 _ / xwo vertices are Im
the results of counterclockwise rotations by | and by ^ of 2 + i about the center
of the hexagon, and two vertices are the results of counterclockwise rotations
by x and by of 8 - 3i about the center of the hexagon. Therefore, the vertices
are:
Then the other vertices of the hexagon are given by
e”i/3[(2 + 0 - (5 - 0] + (5 - 0 = ( i + ^>J (-3 + 20 + (5 - i)
7-2V3 3V3
-1 ,
g2ni/3[(2 + i) _ (5 _ f)] + (5 - i) = -^ + ^ )(-3 + 20 + (5-0 =

i
13-2V3 4 + 3V3.
13 + 2V3 -4 + 3V3.
eni/3[(8 - 30 - (5 - 0] + (5 - 0 = --- + -T-l
I + "f'j(3"2° + (5“!)
7 + 2V3 3V3.
g2™73[(8 _ 30 - (5 - 0] + (5 - 0 = ~\ + "F,j (3-2i) + (5-0 = ---(- —— 1
8.41 Let the vertices of the trapezoid be A, B, C, and D, where the parallel
sides are AB and CD, and let the corresponding complex numbers be a, b,
c, and d, respectively. Without loss of generality, we can orient trapezoid
ABCD so that both a and b are real and b > a. Then AB = b - a.
Since AB || CD, c and d have the same imaginary part, say h. Let
c = p + hi and d = c\ + hi, where p and q are real numbers. Note that
c-d = (q+ hi) - (p + hi) = p - q, so CD = p - q.
Let M and N be the midpoints of BC and AD, so the corresponding complex numbers are
b+c b + p + hi a+d a + q + hi
m - and n =
~Y~ 2 2 2
163
Then
b + p + hi a + q + hi (b-a) + (p-q)
m-n =---r-=---,
2 2 2
which is a real number, so MN is parallel to AB and CD. Furthermore, MN = AB2CD-, as desired.
8.42 Without loss of generality, we can assume that one of the vertices of the paral¬ Im
lelogram is the origin. Then the other vertices of the parallelogram are of the form a, b
b, and a + b.
The diagonals are perpendicular if and only if the line through 0 and a + b is
perpendicular to the line through a and b. This occurs if and only if
Re
((a + b) - 0)(a -b) + (a + b - 0)(a - b) = 0,
which is equivalent to (a + b)(a -b) + (a + b)(a - b) = 0. Expanding the left side gives
aa - ab + ab - bb + aa + ab - ab - bb - 0,
which simplifies as aa = bb, or \a\2 - \b\2, or \a\ = \b\. But two sides of the parallelogram are equal in length if and
only if it is a rhombus.
8.43 We know that z lies on the perpendicular bisector if and only if the line joining z and ^ is perpendicular
to the line joining a and b, which is true if and only if
T-. f a+ b\ . ,. /_ a+ b\
(a -b)[z-—J + (a-b) ( z-— J = 0.
Expanding the products gives
- (a - b)(a + b) , ,(a- b)(a + b)

(a - b)z - --£-- + (a- b)z - --^-- = 0,
and rearranging this gives
(a + b)(a -b) + (a - b)(a + b) aa - ab + ab - bb + aa + ab - ab - bb

(a - b)z + (a - b)z aa - bb.
2 2
8.44 We consider the case in which WXYZ is oriented as shown at right. (We can X
tackle the other possible orientation of WXYZ in essentially the same way.) If ATSZ
is an isosceles triangle, it looks like we must have ZTSZ = 90° and TS = SZ. We have
ZZST = 90° and TS = SZ if and only if T is a 90° rotation of Z about S. Letting s, t, and
z be the complex numbers corresponding to S, T, and Z, we can write our condition
for ATSZ to be an isosceles triangle as simply t - s = (z - s)i (for the case that t is a
90° counterclockwise rotation of z about s). That's a pretty simple target, so let's plow
ahead with complex numbers.
We can make life somewhat simpler by choosing convenient coordinates. We

choose the center of the square to be the origin, and orient the axes so that W corresponds to 1, X to i, Y to -1, and
Z to -i. Since T is the midpoint of WX, we have t = -yh Since S is the midpoint of OY, we have s =
Now, we're all set to finish off the proof. We have z = -i, s = and t = Therefore, we have
(z - s)i = (-z - (-5)) i = 1 + 1 and t - s = - (-i) = 1 + 5, so (z - s)i = t - s. This tells us that t is a 90°
counterclockwise rotation of z about s, as desired. Therefore, A TSZ is an isosceles right triangle.
164
Review Problems
8.45 We could tackle this problem with the regular tools from Euclidean geometry, but we'd have to be very
careful to make sure that our proof covered all possible configurations. Instead, we'll use complex numbers, and
tackle all possible configurations at once.
We place the problem on the complex plane such that X is the origin and W is at 1. Since XW is horizontal,
point Y is the image of a vertical translation of X and Z is the image of a vertical translation of W. Therefore, Y is at
ai for some real constant a, and Z is at 1 + bi for some real constant b. The midpoint of YZ then is ai+^+- = \ + î.
Let this midpoint be M. We wish to show that MW = MX. We have MW = (| + -1 \-\ + a-fi\ and
MX =|| + î|. Since |z| = |z| for any complex number z, we have
1 a + b. 1 a + b. 1 a + b. 1 a + b.
MW = = 1-11 = MX.
~2 + 2 L ~2 + 2-L 2 2 1 2 + 2 1
We could also have noted that M is the result of a vertical translation of the point corresponding to \. Therefore,
M is on the perpendicular bisector of WX, which means that it is equidistant from W and Z.
We have no diagram for this problem because there are two possible configurations: Y and Z can be on opposite
sides of Wi or on the same side of fwL Using complex numbers allows us to tackle both at once.
8.46 Let ABCD be a quadrilateral and let a, b, c, and d be complex numbers corresponding to the respective
vertices of the quadrilateral. Then, the midpoints of AB and CD are +Jp and Lp. The midpoint of the segment
connecting these two points is ^W/.2+_(c+lj)/2 - a+bfc+d _ sj_nce q-qs eqUals LUl/UWW2^ the complex number a+b+c+d
also corresponds to the midpoint of the segment connecting the midpoints of AD and BC.
The midpoints of the diagonals AC and BD are ^ and Lp. The midpoint of the segment connecting these
two points is (g+c)/2+(fr+rf)/2 = a+b+c+d' coincides with the midpoint of the segment whose endpoints are the
midpoints of AB and CD, so the two segments described in the problem bisect each other. (We can go through the
same steps with sides BC and AD, as well.)
8.47 Solution 1: Use Translation. The translation that maps A to B also maps D to C. The translation that maps A
to B maps any complex number z to z + (b - a), since a + (b - a) = b. Therefore, this translation maps d to d + (b - a).
Since this translation also maps D to C, we must have d + {b -a) = c, so d-c + a-b.
Solution 2: Use Diagonals. Since ABCD is a parallelogram, the diagonals must bisect each other. That is, their
midpoints must be the same, so ^p = ^p, from which we find d = a + c-b, as before.
8.48 From the Triangle Inequality, we have \w - z| + |z| > |(w - z) + z| = \w\, so \w - z\ > \w\ - |z|.
8.49 Let p, a, b, c, and d be complex numbers corresponding to P, A, B, C, and D, respectively. So, we must show
that
|a - p\2 + \c - p|2 = \b - p\2 + |d- p|2.
We can get rid of the magnitudes by noting that |z|2 = zz for all complex numbers z, but first we simplify the
problem by choosing our origin wisely. We let the center of the rectangle (the intersection of the diagonals) be the
origin. This point is equidistant from all 4 vertices and is the midpoint of AC and BD. Therefore, we have c = -a,
d = -b, and |fl| = \b\. So, we have
| a - p|2 + |c - p|2 = | a- p|2 + |-a - p\2 = (a - p)(d - p) + (-a - p)(-a - f)

= ad - ap - pa + pp + ad + ap + pa + pp
= 2 (ad + pp).
Similarly, we have |b - p|2 + |d- p|2 = 2{bb + pp). Since |a| = |6|, we have |a|2 = |&|2, so ad = bb and ad + pp = bb + pp.
Therefore, we have |a - p|2 + |c - p|2 = \b- p|2 + |d- p|2, so AP2 + CP2 = BP2 + DP2, as desired.
165
Challenge Problems
8.50 We save ourselves a lot of work by choosing the origin and real axis wisely. Im
We place the problem on the complex plane by letting O be the origin and letting k be
the real axis. Then the distance from each vertex to k is the real part of the complex
number that corresponds to the vertex. We let x + yi correspond to one vertex of the
square. We find another vertex by rotating x + yi by | counterclockwise about the
origin, to get (x + yi)i = —y + xi. The next vertex is a | counterclockwise rotation of
-y + xi about the origin, so it is at (-y + xi)i = -x - yi. The last vertex is a f rotation
of -x - yi about the origin, or (-x - yi)i = y - xi.
The sum of the squares of the distances from each vertex to the real axis is y2 + x2 + y2 + x2 = 2(x2 + y2). The
quantity x2 + y2 equals the square of the distance from any vertex to the origin. This distance equals half the
length of a diagonal of the square, and this length is the same no matter how the square is oriented about the
origin. Therefore, the quantity x2 + y2 is the same no matter how the square is oriented about the origin, which
means that the sum of the squares of the distances from the vertices of ABCD to k is the same for any orientation
of the square about the line k. Equivalently, this sum is the same for any location of k through the center of a fixed
square ABCD.
8.51 From the diagram, we expect that if APAQ is an isosceles right triangle, then
IPAQ = | and PA = AQ, so we want to show that P is the image of a \ rotation of Q
about A. We place the problem on the complex plane, letting the lowercase version
of each letter be the complex number corresponding to the respective uppercase
point. With the diagram oriented as shown, showing that P is the image of a j
counterclockwise rotation of Q about A is the equivalent of showing that p - a =
ent/2(q - a) = i(q - a).
We simplify the problem by letting A be the origin, so a = 0 and we wish to prove

that p = -iq. Now, we must find p and q in terms of b and c, and check if we have
P = -ifi-
In square CAFG, point G is a ^ counterclockwise rotation of A around C, so we have
g-c = em/2(a - c) = i(a - c).
Since a = 0, we have g = c - ic.
Similarly, D is a | clockwise rotation of B about C, so
d- c = e~ni/2(b - c) - (-i)(b - c) - -ib + ic,
which means d = -ib + c + ic.
Finally, we find q from parallelogram CDQG. The diagonals of the parallelogram bisect each other, so ^
which means
q-d + g- c = (-ib + c + ic) + (c - ic) - c = -ib + c.
Similarly, H is a | clockwise rotation of A about B, so h - b = e~m/2(a - b) = (-z)(0 - b) = ib, which means

h = b + ib. Since / is a | counterclockwise rotation of C about B, we have j - b = em/2(c - b) = i(c - b), so
j = b + i(c - b) - (1 - i)b + ic. Parallelogram BHPJ then gives
p = j+h-b = ((1 - i)b + ic) + (b + ib) -b -b + ic.
166
Challenge Problems
We therefore have iq - i(—ib + c) = b + ic - p, so p is a | counterclockwise rotation of q about the origin. Since

A is the origin, we have proved that P is a j rotation of Q about A, which means that A FAQ is an isosceles right
triangle, as desired.
8.52 We find p first. Since a, p, and b are collinear, we have
(p - b)(a -b) - (p- b)(a - b) = 0.
Since the line joining p and 2 is perpendicular to the line joining a and b, we have
(p - z)(a -b) + (p - z)(a - b) - 0.
We can cancel p by adding these two equations. When we add the equations, we group the (a — b) terms and
group the (a - b) terms. This gives us
(2 p-b - z)(a — b) + (b - z)(a - b) = 0.

Rearranging this equation gives
(2p - b - z)(a -b) = (z — b)(a - b).
Since a £ b, we can divide by (a - b), and then isolate p to find
1 / (z - b)(a - b) ^ z(a - b) - ab + bb + ba - bb + z{a - b) (a - b)z + (a - b)z - ab + ab

P= 2 1 a-b 1 ~ 2{a - b) 2(a - b)
(1a - b)z - ab + ab
Since p is the midpoint of z and w, we have p = ^. Solving for w gives vo -2p - z-
a—b
8.53 The expressions + (x - l)2 and y/x2 + (x + l)2 resemble the distance formula. We are further inspired
to think of them in this way by being asked to minimize their sum, because minimizing a sum of distances
reminds us of the Triangle Inequality, AB + BC > AC. But what points should we take to be A, B, and C to give
the expressions in the problem? Taking these to be points in the complex plane, we can let B represent x + xi, A
represent i, and C represent -i. Then, we have
AB + BC= a/x2 + (x - l)2 + \Jx2 + (x + l)2.
By the Triangle Inequality, this sum is minimized when B is on AC, which is a fixed segment that does not depend
on x. Since AC is the segment from i to -i, point B must be on the imaginary axis to minimize AB + BC. The only
point of the form x + xi on the imaginary axis is 0, so the minimum value of / is /(0) = [Tj.
Looking back over our geometric argument, we can now construct a purely algebraic argument to show that
the minimum value of / is 2. Let z = x + xi, a = i, and b = -i. Then
|z — a\ + |z — b\ - |(x + xi) - i\ + |(x + xi) + z]
= |x + (x - l)z'| + |x + (x + l)z|
= a/x2 + (x - l)2 + a/x2 + (x + l)2,
which is /(x). By the Triangle Inequality, we have \z - a\ + |z - b\ = |a - z| + |z -b\> \(a - 2) + (z - b)\ = \a - b\ =

|i - (-z')| = |2z'| = 2, so /(x) > 2 for all x. Furthermore, /(0) = VO2 + l2 + VO2 + l2 = 1 +1 = 2, so the minimum value
of /(x) is [2].
8.54 We present two solutions—one that is largely geometric and the other largely algebraic.
167
Solution 1: Geometry. The inequality \z\ - 5 + 3z'| < 4 means that

Z\ lies inside or on the circle with center 5 - 3z and radius 4, and the
inequality \z\ - 5i\ < 2 means that Z2 lies inside or on the circle with
center 5z and radius 2. The distance between the centers of the circles
is |(5z) - (5 - 3z)| = V89, and the sum of their radii is 6. Since 6 < y/89,
the circles do not intersect.
Let the line joining 5 - 3z and 5z intersect the two circles at A, B, C,

and D, as shown at right. The expression |zi - Z2I equals the distance
between z\, which is in or on the larger circle, and Z2, which is in or
on the smaller circle. We claim that the shortest distance between two
points with one point on each circle is BC and the longest distance is
AD.
To see why BC is the shortest possible distance, consider the line

through B tangent to the circle centered at 5z and the line through
C tangent to the circle centered at 5 - 3z. (These lines are dotted in the diagram.) These two tangents are
perpendicular to BLB so they are parallel to each other. Neither circle passes through a point that lies between the
two tangents, so |z2 - Zi| is at least as great as the distance between the two tangents, which is BC.
To see why AD is the largest possible distance, consider the circle with diameter AD. (This circle is dotted in
the diagram.) The circle with diameters AB and CD are internally tangent to the circle with diameter AD, so, other
than points A and D, both circles are entirely inside the large circle. Therefore, |Z2 - zi | can be no larger than the
length of the diameter of the large circle, which is AD.
To find the lengths of BC and AD, we recall that the distance between the centers of the smaller circles is V89-
Then, BC is this distance minus each of the radii of the circles, and AD is this distance plus each of the radii of the
circles. That gives us BC = V89-2-4 = V89-6 as the minimum value of |î —Z2I and AD = V89+2+4 = 6 + V89
as the maximum value of \z\ - Z2I
Solution 2: The Triangle Inequality. This solution is a little less intuitive, but our first solution helps guide us.
First, we'll tackle the maximum value. We wish to show that |zi - Z2I < k, where k is some constant. To use the
Triangle Inequality, we want to break zi - Z2 into a sum of expressions that we know something about. We know
that |zi - 5 + 3z| < 4 and |z2 - 5z| < 2, and we can write the second inequality as |5z - Z2I <2. But Z\ - Z2 is not exactly
the sum of zi - 5 + 3z and 5z - Z2, so we'll need one more term, (5 - 3z) - 5z, and we have
Zi - z2 = (zi - 5 + 3z) + ((5 - 3z) — 5z) + (5z - Z2).
We have a hint that we're barking at the right tree when we notice that the middle term on the right is the difference
of the complex numbers corresponding to the centers of the circles in our first solution.
Apply the Triangle Inequality gives
|zi - Z2I = |[zi - 5 + 3z] + [(5 - 3z) - 5z] + [5z - z2]| < \z\ - 5 + 3z| + |(5 - 3z) - 5z| + |5z - z2|.
We know that \z\ - 5 + 3z| < 4, |(5 - 3z) - 5z| = |5 - 8zj = V89, and |Z2 - 5z| < 2, so we have
|zi - Z2I < |zi - 5 + 3zj + |(5 - 3z) - 5z| + |5z - Z2I < 4 + V89 + 2 = 6 + V89.
We can show that the maximum can be achieved in the same way we did in the first solution. We consider the
circles that are the graphs of |z - 5 + 3z| = 4 and |Z2 - 5z| = 2, and take point D on the first circle and A on the second
such that the centers o both circles are on AD. The distance between the centers is V89, and the sum of the radii
is 6, so AD - 6 +V89
Finding the minimum is a little trickier, because here, we want |zi — Z2I > k, where k is some constant. To apply
the Triangle Inequality to produce |zi - z2| on the greater side, we'll have to add more expressions to it. But we
168
Challenge Problems
have some candidates to try. We know that \z\ — 5 + 3z'| < 4 and |z2 - 5i\ < 2. This time, we "reverse" the terms
in the first inequality and write it as |5 - 3z - zi | < 4, where we do this so that Z\ will cancel when we apply the
Triangle Inequality:
|5 - 3i - zi| + |zi - z21 + |z2 - 5*| > |(5 - 3* - za) + (z1 - z2) + (z2 - 5z)| = |5 - 3* - 5i\ = V89.
Therefore, to minimize \z\ - z2|, we wish to maximize the sum of the other two terms on the left above, which are
|5 - 3z - zi| and |z2 - 5z|. We can use the given inequalities to write
|5 - 3* - zi| + \z\ - z2| + |z2 - 5*| < 4 + |zi - z2| + 2 = 6 + |zi - z2|.
Putting this together with the Triangle Inequality result, we have
6 + |zi - z2| > [5 - 3* - Zi| + |zi - z2| + |z2 - 5*| > V89.
Taking the two ends of this, we find |zi - z2| > -6 + V89. We can show that the minimum can be achieved by
considering the circles that are the graphs of |z - 5 + 3z| = 4 and |z2 - 5z| = 2. We take point C on the first and B on
the second such that the segment connecting the centers of both circles contains B and C. The distance between
the centers is V89, and the sum of the radii is 6, so BC -6 +V89
8.55 Let a = 2, b = 1 + 2z, c = -1 + z, and d = -1 - z, so |z - 2| + |z - (1 + 2z)| +

|z - (-1 + z)| + |z - (-1 - z)| = \z-a\ + \z-b\ + \z-c\ + \z- d\.
By the Triangle Inequality, we have
\z - a\ + \z - c\ - \a - z\ + \z - c\ > \(a - z) + (z - c)| = \a - c\ = |3 - z| - VlO,
and
Re
\z-b\ + \z-d\ = \b-z\ + \z-d\> \{b - z) + (z - d)\ = \b-d\ = \2 + 3z'| = Vl3.
Furthermore, equality occurs in both inequalities when z is on the segment

from a to c and on the segment from b to d. Therefore, we will have equality at
the point at which these segments intersect, if they do intersect. The diagram above shows that they do indeed
intersect (we could also show this algebraically), so the minimum value is VlO +Vl3
8.56
(a) Solution 1: Apply Vieta's Formulas for quadratics. Let Ci and C2 be the two possible locations Ci
of C such that A, B, and C form an equilateral triangle, as shown at right, and let C\ and c2 be
A
the complex numbers corresponding to C\ and C2.
Let C be a primitive sixth root of unity. Since C\ is the image of a | counterclockwise
rotation of B around A, we have
ci = C(b - a) + a = (1 - C)a + Cb. C2
Also, C2 is the image of a | counterclockwise rotation of A around B, so
C2 = C(« - b) + b = Ca + (1 - Qb.
The equation we wish to prove true can be viewed as a quadratic in c. This equation has the roots Ci and
c2 given above. We know that if a quadratic has the form x2 +px + q, then the sum of the roots is -p and the
product of the roots is q. So, we can find the quadratic in c with roots Ci and c2 by finding Ci + c2 and CiC2.
169
First, we have
C\ + C2 — (1 — C)a F Q> + C,u + (1 — C,)b — a + b,
Then, as explained in the text, C is a root of x2 - x +1, so C - C +1 =0, which means C = £ - L and we have
cic2 = [(l-Ofl + C&][Cfl + (l-C)&]

= (C - C2)fl2 + (1 - Q2ab + Cab + (£ - C)b2
= (C - CV + (1 - 2C + C2)flb + Cab + (c - C)b2
= (C - C V + (1 - 2C + 2C)ab + (£ - C)b2
= (£ - c + l)fl2 + (1 - 2£ + 2£ - 2)flfc + (£-£ + 1)&2
= a2 - ab + b2.
Therefore, as explained above, Ci and C2 are the roots of
c2 - (a + b)c + a2 -ab + b2 - 0,
which rearranges as a2 + b2 + c2 = ab + ac + be.

Solution 2: Apply Vieta's Formulas for cubics. Let f(x) = (x - r)(x - s)(x - t). Expanding the right side gives
f(x) = x3 - (r + s + f)x2 + (rs + sf + tr)x - rst.
Therefore, the sum of the roots of a cubic of the form x3 + a^x2 +a\X + ao is -02, the sum of all three products
of pairs of roots is a\, and the product of all three roots is -«o- These are Vieta's formulas for a cubic.
To apply them to this problem, consider the cubic f{x) - (x-a)(x- b)(x-c), where a, b, and c are complex
numbers. The roots of f(x) correspond to the vertices of the triangle formed by connecting a, b, and c in
the complex plane. Let this be aABC. We'd like to relate / to a cubic we know a lot about, and preferably
a cubic whose roots are the vertices of an equilateral triangle in the complex plane. A natural choice is
g(z) - z3 - 1, whose roots are the cube roots of unity, which are the vertices of an equilateral triangle in the
complex plane. Let this triangle be APQR.
Triangle ABC is equilateral if an only if we can map it to A PQR by a translation followed by a dilation,
followed by a rotation. First, we translate A ABC such that its centroid is the origin. Then, we dilate the
triangle about its center by a factor of where r is the circumradius, and finally, we rotate the triangle
about the origin such that one of the vertices is at 1 in the complex plane. In each transformation, the
original triangle and its image are similar, so A ABC is equilateral if and only if the result of these three
transformations is. Let's see what the effect of these transformations is in complex numbers.
The translation maps x to x-t, where t is a complex number. The dilation maps x - t to k(x - t), where
k is a nonzero real number. The rotation maps k(x - t) to cok(x - t), where co is complex number such that
M = 1.
Therefore, we can map the roots of f(x) = (x - a)(x - b)(x - c) to the roots of g(z) = z3 - 1 with the
transformation z = cok(x - t) = cokx - cokt, where z is a root of g(z) and x is a root of f(x). Letting u = cok and
v = -cokt, we have z = ux + v for some complex numbers u and v. Substituting this into our expression for
g gives
- n3
3 3v
uu 2 3v2 v3 1
g(ux + v) = (ux + vf - 1 = u3x3 + 3u2vx2 + 3uvzx + v6 - 1 = u X H-X + —~X + —r--
U U1 Ud U3
Since this transformation maps A ABC to A PQR, the polynomials f(x) and g(ux + v) have the same roots.
Furthermore, since / is monic, we have
1 3 3v 2 3v2 v3 1
/(x) = xg(uX + V) = X5 + —XL + —r-x + -=-r.
M3 U U2 U3 U3
170
Challenge Problems
Applying Vieta's formulas for a cubic, the sum of the roots of the cubic in x must be and the sum of
pairs of these roots must be . Since the roots of / are a, b, and c, we have
a + bu + c =-,
u
Squaring the first equation, and noting that writing yr = 3 • allows us to substitute ab + bc + ca for ^r,
gives
u2 u
Expanding (a + b + c)2 gives a2 + b2 + c2 + 2(ab + be + ca) = 3(ab + be + ca), so a2 + b2 + c2 = ab + be + ca, as
desired. Therefore, the triangle is equilateral if and only if a2 + b2 + c2 = ab + be + ca.
(b) Multiplying both sides by (a - b)(b - c)(c - a), the desired equation becomes
(b - c)(c -a) + (c- a)(a -b) + (a - b)(b - c) = 0

o be - ab - c2 + ac + ac - be - a2 + ab + ab - ac - b2 + be = 0
o ab + ac + be — a2 - b2 - c2 - 0
<=> a2 + b2 + c2 - ab + ac + be,
which is the same equation as in part (a).

(c) Expanding, the left-hand side becomes
(b - c)2 + (c - a)2 + (a - b)2 = b2 - 2be + c2 + cz - lac + a2 + a2 - lab + b2 = 2a2 + lb2 + lc2 - lab - lac - 2bc = 0,
which simplifies as a2 + b2 + c2 = ab + ac + be. This is the same equation as in part (a).
8.57 Since Sn is the minimum value of the sum, we have
>/(2 -1 - l)2 + a2 + yj(2-l-l)2 + a2 + ■■■+ y/(2-n- l)2 + a2 > Sn
for all n. An inequality with a sum on the greater side—that looks like the Triangle Inequality. If we write each z,-
in
|Zll + lz2l -I-+ \zn\ > |Zl + Z2 +-1 Z„|
Applying this to the sum of square roots in our problem, we have
\J{2 ■ 1 — l)2 + a2 + yj{2 ■ 2 - l)2 + a\ + ■ • ■ + yj{2 ■ n - l)2 + a2 >

\{1 • 1 - 1) + (2 • 2 — 1) + • • • + (2 • n — l)p + (fli + £?2 ■ ■ ■ + An)2-
The suma\+ a2 + ■■■ + an is given as 17, and
(2 • 1 - 1) + (2 ■ 2 - 1) + • • • + (2 • n - 1) = 2(1 + 2 + • • • + n) - n = 2 ■ -n = n2,
171
so the inequality now is simply
y/(2 • 1 — l)2 + a\ + y/{2 ■ 2 - l)2 + a\ + • • • + y/(2 ■ n - l)2 + a2n > Vn4 + 172.
Now, our problem is to find an integer n such that S„ = Vn4 + 172 is an integer. We square both sides to get rid
of the square root and find S2 = n4 + 172. Subtracting n4 from both sides and factoring S2 - n4 as a difference of
squares gives (S„ - n2)(Sn + n2) = 172. We know that S„ + n2 is positive, so Sn - n2 must be positive as well. Both of
these expressions must be integers, so we have only two options: either Sn + n2 = S„ - n2 = 17, or Sn +n2 = 172 and
Sn -n2 = 1. The former possibility gives Sn = 17 and n = 0, but n must be positive. The latter gives Sn + n2 = 289
and Sn - n2 = 1, which means S„ -- 145 and n = 12
8.58 We let the origin be the point at which the four given lines meet, and we let a, b, c, d, e be complex numbers
corresponding to the respective vertices in the diagram. Because the origin is on the line through A and the
midpoint of EB, we have
(o-^)(0-3)-(o-f±l)(0_a) = o.
Simplifying this equation gives

ba + ea- ab - ae = 0.
Similarly, the other three given lines pass through the origin and yield
cb + ab - be - ba = 0,
dc + bc -cd-cb = 0,
ed + cd- de - dc - 0.
Adding all four of these equations gives a whole lot of cancellation, and leaves
ea + ed-ae- de = 0. (8.2)
We'd like to show that the line through E and the midpoint of AD passes through the origin, which is true if
and only if
/ a + d\ - f- a + d\
(0 - -y~J (0 -e) - f 0-— J(0-e) = 0.
Simplifying this equation gives

de + de - ed — ed = 0.
Multiplying Equation (8.2) by -1 gives us this desired equation, so the line through E and the midpoint of AD
passes through the origin.
8.59
(a) Let co - e2n,/n be a primitive nth root of unity, so con = 1. Without loss of generality, we can assume that the
point Pjt corresponds to the complex number cek in the complex plane, 0 < k < n - 1. Let z be the complex
number represented by P, so |z| = 1. Because co is an nth root of unity, we have cok = co~k = con~k. Then
PP\ = \z- cok|2 = (z - cok)(z - cok) = (z - cok)(z - co~k) = (z - cok)(z - con~k)
= zz - con~kz - cokz + con - |z|2 - con~kz - cokz + 1=2- con~kz - cokz,
172
Challenge Problems
so
pp20 + pp2 + pp2 + ...+ pp2_l
= (2 - conz - z) + (2 - ojn~xz - coz) + (2 - oon-2z - cv2z) + ■ • • + (2 -coz- con~lz)
= 2n - (co + oj2 + at3 + ■ ■ ■ + con)z - (1 + co + co2 + • ■ • + o;”_1)z
= 2n - (co + co2 4-+ con~l + l)z - (1 + cl> + ce2 H-+ con~l)z
= 2n — (1 + co + co2 + • • • + con~1)z - (1 + co + co2 + ■ ■ ■ + con~l)z
By the formula for a geometric series, we have 1+<x>+cl>2-I—+con~1 - = 0, soPP2+PP?+PP?+- • -+PP2 n =

r\ I—Cl) ' U 1 Z Tl— I
(b) From the equation PP2 = 2- con~kz - cokz, we have
PPt = (2 - con~kz - cokz)2 = 4 + co2n~2kz2 + o;2fcz2 - 4a;-fcz - 4cokz + 2cvnzz
- 4 + ce2" 2Az2 + co2kz2 - 4con~kz - 4a/z + 2 = 6 + cu2n~2kz2 + ce^z2 - 4con~kz - 4cokz.
Then
PPo + PPt + PPt + • • • + = (6 + a;2"z2 + z2 - 4<d"z - 4z)

+ (6 + co2n~2z2 + co2z2 - 4con~lz - 4cvz)
+ (6 + o)2"_4z2 + co4z2 - 4a;n_2z - 4co2z)
h-1- (6 + co2z2 + coZn~2z2 - 4coz - 4con~lz)
= 6n + (co2 + co4 + co6 + • • • + co2n)z2 + (1 + co2 + co4 + • ■ ■ + co2n~2)z2
- 4(co + a;2 + co3 + • • • + con)z - 4(1 + co + co2 + • • • + con~l)z
= 6n + co2( 1 + co2 + oj4 +-b ci)2"_2)z2 + (1 + a>2 + co4 + ■ • + co2n-2\—2

)ZZ
- 4a)(l + co + <vz + • • ■ + con x)z - 4(1 + co + a>2 + • • • + conn-l\=
l)z,
where we used con = 1 and co2n = (con)2 = 1 in the last step. We know that 1 + co + co2 -\-+ con_1 = 0, and
n\2
1-C02n _ 1 - (conj
1 + 6J2 + Cl)4 + • • • + Ci)2" 2 = = 0,
1 - O)2 1 - Cl)2
so PP40 + PP2 + PP$ + --- + PP4_1 = 6n.
8.60 We could let z8 = x + yz and go through some clever algebraic manipulations to Im

find the possible values of z28 and z8, but a little geometric insight offers a significant
shortcut. Rearranging the given equation as z28 = z8 + 1, we see that z28 is the image
of a translation of z8 by 1 unit to the right. We also know that |z| = 1, so |z28| = |z8| = 1
as well, which means that z28 and z8 are on the unit circle. There are only two pairs Re
of points on the unit circle such that one point is 1 unit to the right of the other. These
are shown at right.
To see that these are the only two possibilities, we note that there are only at
most two points on the unit circle with a given y-coordinate. The difference in the
x-coordinates of these two points is 2 when the y-coordinate is 0, and the difference strictly decreases as y moves
away from 0. Therefore, this difference can only equal 1 once above the x-axis and once below the x-axis, and
these are at the values shown in the diagram.
Since z28 is the result of translating z8 by 1 unit to the right, the two possibilities are z8 = e2m/3 and z28 = emf3,
and z8 = e4m/3 and z28 = e5n^3.
173
If z8 = g2™/3 and z28 = enl/3, then z24 = (z8)3 = (g2™/3)3 = e2n‘ = 1, and
-r28
Z = ,24
= eni/3.
The solutions to this are of the form z = emll2+knl2/ where k = 0,1, 2, and 3. These arguments correspond
to the angles 15°, 105°, 195°, and 285°. Furthermore, if z4 = em/3, then z8 = (z4)2 = (em/3)2 = e2m/3, and
z28 = (z4)7 = (em/3)7 = e7m/3 = e™/3, so all these solutions work.
If z8 = e4m//3 and z28 = e5m/3, then z24 = (z8)3 = (g4ra/3)3 = g12ra = 1, and
728
74 _ £_ _ _5ra'/3
Z - Z24 - C •
The solutions to this are of the form z = e5mP2+kn/2/ where k = 0, 1, 2, and 3. These arguments correspond to
the angles 75°, 165°, 255°, and 345°. Furthermore, if z4 = eSTn/3, then z8 = (z4)2 = (g5™/3)2 = g10m/3 - e^m/3f and
Z28 - (z4)7 = (g5ra’/3^7 _ e35ni/3 _ g57i//33, so all these solutions work.
Therefore the angles 0, in increasing order are 15°, 75°, 105°, 165°, 195°, 225°, 285°, and 345°, and the sum we
seek is 75 + 165 + 225 + 345 = 840
8.61 The center of the decagon is 2 (as a complex number), in the com¬
plex plane. Moreover, the decagon is a rightward translation by 2 units
of the decagon whose vertices are the tenth roots of unity. Hence, each
complex number x^ + yî is of the form 2 + z, where z is a 10th root of unity. Re
Let co = g2™/10. Then
{x\ + y\i)(x2 + y2i)(%3 +1/30 • • • (*io + yioO = (2 +1)(2 + cv)(2 + co2) • • • (2 + co9).
The 10th roots of unity are the roots of xw -1 = 0, so
(x - l)(x - co)(x - co2) ■■•(x- co9) = xw - 1.
Taking x = -2, we get (-2 - l)(-2 - co)(-2 - co2) • ■ • (-2 - co9) - (-2)10 - 1, which simplifies as
(2 + 1)(2 + co)( 2 + co2)---( 2 + co9) = 1023
8.62 The area of the equilateral triangle is maximized when one vertex of the triangle Im
coincides with a vertex of the rectangle, and the other two vertices of the triangle lie
on the sides of the rectangle, so let the vertices of the equilateral triangle be A, P, and
Q, where P lies on BC and Q lies on CD.
We place the rectangle in the complex plane so that A corresponds to the origin
and C corresponds to 11 + 10/. Let p and q be the complex numbers corresponding
to P and Q, respectively, so that the real part of p is 11 and the imaginary part of q is
10. Let p = 11 + xi and q - y + 10/.
Then q can be obtained by rotating p by | counterclockwise about the origin, so q = ent/3p, or
n . . n 1n. 11 x. llV3. xV3

y +10/ = ( cos — + i sin — ) (11 + xi) y + 10i =(\ + ~yi) (H+x/) y + lOz = — + -M - i —.
3 3
174
Challenge Problems
Equating the real parts and equating the imaginary parts, we obtain the system of equations
11 V3
y = ~2 ~ ~YX'
1 11V3
10 = -x + —„—
From the second equation, we have x = 20 - 11 V3. Then
|p|2 = |11 + xz'|2 = ll2 + x2 = 121 + (20 - 11 V3)2 = 884 - 440 V3,
so the area of triangle APQ is ^ |p|2 = (884 - 440 y/3) = -330 + 221 V3
8.63 Let co — e2™^, a primitive cube root of unity, so co1 + co + 1 = 0. Without loss of generality, let the vertices of
the equilateral triangle in the complex plane be 1, co, and co2, so s = V3.
Let p be the complex number corresponding to the point P. Then in some order, PA2, PB2, PC2 are
IV ~ 1|2 = (p- 1 )(p ~ 1) = Ipl2 -p-p + 1,

IP - co\2 = (p - co)(p -co) = (p- co)(p - co2) = |p|2 - co2p -cop + l,
| p - coz\2 - (p - co2)(p - co1) = (p~ co2)(p — co) — \p\2 - cop- co2p + 1.
Then
PA2 + PB2 + PC2 + s2 - (|p|2 -p-p + 1) + (|p|2 - co2p -cop + 1) + (|p|2 -cop- co2p + 1) + s2
= 3|p|2 - (1 + co + co2)p - (1 + co + co2)p + 6
= 3|p|2 + 6,
so (PA2 + PB2 + PC2 + s2)2 = 9|p|4 + 36|p|2 + 36.
Also, we have
PA4 + PB4 + PC4 = (|p|2 - p - p + l)2 + (|p|2 - co2p cop + l)2 + (|p|2 -cop- co2p + l)2
= |p|4 + p2 + p2 + 1 - 2|p|2p - 2|p|2p + 2|p|2 + 2|p|2 - 2p - 2p
+ |p|4 + cop2 + co2p2 + 1 - 2ra2|p|2p - 2ca|p|2p + 2|p|2 + 2|p|2 - lco2p - 2cop
+ |p|4 + co2p2 + cop2 + 1 - 2cr»|p|2p - 2w2|p|2p + 2|p|2 + 2|p|2 - 2cop - 2co2p
= 3|p|4 + (1 + co + co2)p2 + (1 +co + co2)p2 + 3-2(1 + co + co2)p -2(1 + co + co2)\p\2p
- 2(1 + co + o>2)|p|2p + 12|p|2 - 2(1 + co + co2)p -2(1 + co + co2)p
= 3|p|4 + 12|p|2 + 3,
where we used the fact that 1 + co + co2 = 0 to eliminate many terms in the last step. We therefore have
3 (PA4 + PB4 + PC4 + s4) = 3(3|p|4 + 12|p|2 + 3 + 9) = 9|p|4 + 36|p|2 + 36,
so 3 (PA4 + PB4 + PC4 + s4) = (PA2 + PB2 + PC2 + s2)2.
175
8.64 We know how to handle equilateral triangles and testing for perpendicular¬ Z
ity on the complex plane, so we place the problem on the complex plane. Let C be
a primitive sixth root of unity and let the lowercase version of each letter be the
complex number corresponding to the respective uppercase point.
From equilateral triangles BCX, CAY, and ABZ, we have
x = C(b - c) + c - Cb + (1 - Qc,
y = C(c - a) + a - Q + (1 - Qa,
z = Qa - b) + b = Ca + (1 - Qb,
so
a+z (1 + Qa + (1 - Qb
m -
~2“ 2
c+x Cb + (2-Qc
n =
~Y~ <-i /
b+c
v =
c+y (1 - Qa + (1 + Qc
n
2 2
so
(1 + Qa + (1 - 2Qb + (-2 + Qc
and
, ,, (1 -Qa-b + Qc
n _b =---.
We have B'N' J_ MN if and only if (m - n)/(n' - b') is imaginary. We can make our work a little easier by
choosing A to be the origin, so a = 0, and we have
m-n _ (1 - 2Qb + (-2 + Qc

n' -b' —b + Cc
Uh-oh. That expression on the right doesn't look imaginary. The fact that it still has b and c in it is a problem,
since b and c can be anything. So, we wonder if this expression is equivalent to an expression that doesn't have b
or c. This will be the case if the ratio of the coefficients of b equals the ratio of the coefficients of c, which will give
us some convenient cancellation. Let's check if that's the case by comparing to =^. We have
^ = C-2C2 = C-2(C-1) = 2-C

2-C 2-C 2-C ‘
Sure enough, they're the same. So, there's a constant k such that = k, and we have
m-n _ (1 -2Qb + (-2 + Qc _ k(-l)b + k(Qc _ -b + Cc

n' -b' —b + Cc -b + Cc -b + Cc
Now, we just have to show that k is imaginary. We have k = — 1 + 2C, and the only possibilities for C are the
primitive sixth roots of unity, which are \ + and \ - ~i. For each of these values of C, the value ofk = -1 + 2C
is indeed imaginary. We have therefore proved that is imaginary, so B'N' _L MN.
176
CHAPTER 9 i
l Vectors in Two Dimensions
9.1.1 Let v = Then, we have
0+ V\
0+v + = V,
0+ V2
0v\ 0
Ov = 0 = 0.
0V2 0
9.1.2
4
(a) = y/42 + (~5)2 = V41
(b) ||7i - 24j|| = ^T2 + (-24)2 = V625 = \25
9.1.3
-1 4 \ (10 -1 15
(a) -4s)-(-53 7 -3) + (-6 7 -16
-1 -12 0 11
(b) 4 12 -10
6
9.1.4 The vector AB represents the vector pointing from A to B, and the vector CD represents D
the vector pointing from C to D. If these two vectors are equal, then AB and CD are equal in C
length and parallel, so quadrilateral ABDC is a parallelogram.
B
A
9.1.5 Let v = ( Then cv = , so
cx
||cv|| = = yj (cx)2 + (cy)2 = yjc2x2 + c2y2 = \J c2(x2 + y2) = V? yjx2 + y2 = |c|||v||
cy
177
CHAPTER 9. VECTORS IN TWO DIMENSIONS
9.2.1
-2
(a) = 5 • (-2) + (-3) •4 = -10 + (-12) = [-22
(b) = (-6) • 0 + (-1) • (-3) = 0 + 3 = [|].
9.2.2 We have u - v = (5i - 12j) - (3i + 4j) = 2i - 16j, and u + v = (5i - 12j) + (3i + 4j) = 8i - 8j, so
(u - v) • (u + v) = (2i - 16j) • (8i - 8j) = 2 • 8 + (-16) • (-8) = 16 + 128 = 144
We can also distribute:

(u-v)-(u + v) = u-(u + v)-v-(u + v) = u- u + u- v-(v-u + v-v) = u- u + u- v~u-v-v-v
= u- u — v-v = [52 + (-12)2! - (32 + 42) = 169 - 25 = 144
9.2.3 If the vector is orthogonal to f ^ then (Xj ' \ 2 ) = S^ves + 2y = 0. If x = 5, then
10
y = x/2 = 5/2. If y = 5, then x = 2y = 10. Hence, possible vectors are and
5
u
9.2.4 The vectors p = and q = are orthogonal if and only if p • q = su + tv - 0. Computing
this value for each pair of vectors, we find that the only pairs of orthogonal vectors are , and
f-5
9.2.5 We have AB
-3 i)'.“d AC=h5 . Then
AB-AC 1 • (-5) + (-8) • (-1) -5 + 8

cos A =
||+B||||+C|| s/v + (-8)2 \/(-5)2 + (-1)2 V65V26 13 VlO'
so/l = arccos(i5ij5) »[8?
-5 1 1 -1
Similarly, we have BC , so
4 -3 76) and = (5 -3
(-6)-(-I)+ 7-8 6 + 56 62
cosB =
^(-6)2 + 72 7(-l)2 + 82 V85V65 5V22l'
and B = arccos (d|=) ~
-5 6
The vectors emanating from C are CA = ^ f) and CB = , so
4 -7
5 • 6 + 1 • (-7) 30-7 23
cosC =
V52T12 ^62 + (-7)2 V26V85 V2210'
178
Section 9.3
and C = arccos (^==) 61c
9.2.6
(a) We have v\ = rv cos 0V and v2 = rv sin 0V, so
IMI - \Jv\+v§ - (rv cos 0V)2 + (rv sin 0V)2 = y/r\ cos2 0V + r2 sin2 0V = \/r2(cos2 0V + sin2 0V) = a/^ = rv.
Similarly, ||w|| = rw. Therefore, ||v||||w|| cos = rvrwcos(0v - 0W)-
(b) We also have W\ = rw cos 0W and w2 = rw sin 0W, so by the angle difference formula for cosine,
v\W\ + v2w2 = (rv cos 0v)(rw cos 0W) + (rv sin 0v)(rw sin 0W)
= rvrw(c os 0V cos 0W + sin 0V sin 0W)
= rvrw cos(0v - 0W).
Combining this with part (a), we have
v w = v\W\ + v2io2 = ||v||||w|| cos(0v - 0W).
9.3.1 We have * = 3 - 4s, y = -1 + 5s. Multiplying the first equation by 5 and the second by 4 gives 5x = 15 - 20s,
4y = —4 -t 20s. Adding these two equations gives 5x + 4y = 11
9.3.2 Solving for y gives y = 3x - 4. Letting x = 0 gives y = -4. So, taking t - x as a parameter, the graph of
3x - y = 4 is the same as the graph of = (_^4) + t [^) . There are infinitely many other possible solutions.
To test your solution of the form [ i J = v + fw, make sure the point corresponding to your v satisfies 3x - y = 4,
1
and make sure your w is some multiple of ( ^
9.3.3 As explained in the text, two vectors in are linearly dependent if and only if one is a multiple of the
4
other. Using this condition, we find that the only pairs of linearly dependent vectors are and
-6
(a) We know that at least one such pair (a, b) exists, since every point (x, y) is in the graph of ^ j = au + bv. We
will show that it is impossible for there to be two such pairs.

Let «iu + b\v = w and a2u + b2\ = w, where {a\, bi) and (a2, b2) are distinct pairs of nonzero scalars, so
fliu + &iv = a2u + b2\. Hence,
(fli - a2)u + (b\ - b2)v = 0.
Since u and v are linearly independent, we must have a\-a2-b\-b2- 0, so a\ = a2 and b\ = b2. Therefore,
the pair (a, b) such that w = cm + bv is unique.
179
(b) This part follows immediately from the previous part, which tells us that it is impossible for there to be
two pairs (a, b) if u and v are linearly independent. Therefore, if two different such pairs exist, then u and v
must be linearly dependent.
9.4.1
u•v (2)(3) + (0)(4)

(a) We have projvu = rV
(V32 +42)2
U • V (—3)(2) + (l)(-2)
(b) We have projvu = V =
(y/(2)2 + (-2)2)
9.4.2
(a) As explained in the text, projvu depends on the direction of v, but not on the magnitude of v. Since w has
the same direction as v, the projection of u onto w is the same as the projection of u onto v. The fact that w
has three times the magnitude of v is irrelevant. Therefore, projwu 9i-3j
x•v (3u) • v u• V
(b) We have projvx = —~ v = 2——v = 3 3projvu = 27i - 9j . Geometrically speaking, this
shows that tripling the magnitude of a vector triples the magnitude of the vector's projection onto any
other vector.
9.4.3
(a) Let (x, y) be a point on the line. Since 5i - 3j is normal to the line through (x, y) and (3,4), it must be
orthogonal to the vector (x - 3)i + (y - 4)j. Therefore, we must have (5i - 3j) • ((x - 3)i + (y — 4)j) = 0, so
(5)(x - 3) + (—3)(y - 4) = 0. Expanding and rearranging gives 5x - 3y - 3 = 0
Alternatively, because 5i - 3j is normal to the line, we know the line is the graph of the equation
5x - 3y + c = 0 for some constant c. Since (3,4) is on the line, we must have 5(3) - 3(4) + c = 0, so c — -3.
Therefore, the line is the graph of 5x - 3y - 3 = 0
(b) As explained in the text, the vector 2i+4j is normal to the graph of 2x+4y = 7. But this vector has magnitude
V22" + 42 = 2 V5. We seek a vector with magnitude 5, but with the same direction as 2i + 4j. Since 2i + 4j has
magnitude 2 V5, we can multiply this vector by V5/2 to get the desired vector, V5i + 2 V5j . We also could
have multiplied by - V5/2 to produce - V5i - 2 V5j , which also satisfies the problem.
9.4.4 For any vector w = w\ \ + W2), we have
w-i. (wi)(1) + («72)(0).

projjW = 1 = 1 = w il.
l2
Similarly, we have projjW = w2. Since projxu = projxv for all x, it must hold for x = i and x = j. The equation
projjU = projjV tells us that the first components of u and v are the same, and projjU = projw tells us that the
second components of u and j are same. Combining these observations gives us u = v.
180
Section 9.4
9.4.5
(a) No . We have
(-w) -v w•V
projv(-w) = V = - y V = -projvw.
I|v|| ||V|,
If we also have projv(-w) = projvw, then projvw = -projv w, which means that projvw = 0. All this tells us
is that w is orthogonal to v. If w is orthogonal to v, then both w • v and (-w) • v = 0, so the projections of w
and -w onto v are both 0.
(b) No . Geometrically speaking, the projection of w onto a vector is determined by finding the foot of the
perpendicular from the head of w to the line through the origin that contains the vector from the origin
onto which we are projecting w. This line is the same when we project w onto v as it is when we project w
onto -v. So, the projection is the same in both cases.
Algebraically speaking, we have
w-(-v) w-v w • V
proj(_v)w = -m-mm2 (-v) = —^y(-v) (v) = projvw
! -V
for any vectors w and v.
9.4.6
(a) We have u • ux = (x)(-y) + (y)(x) = 0, so the two vectors are orthogonal. Therefore, the angle between them
is 7Z/2
(b) Let OX and OY be the projections of a onto u and ux. Since u and ux are orthogonal,
we have lOYA = lOXA = ZXOY = 90°, which means that OYAX is a rectangle.
—^ ^ )
Every rectangle is also a parallelogram, so we have A = X + Y. This gives us

a = projua + proj(uX)a.
u•a
Algebraically, we note that projua =-2 u = (u ' a)u because u is a unit vector.
u
Letting a = a\i +112], we have a • u = a\X + ajy, so
projua = (a\x + a.2y){xi + yj) = (a\X2 + ci2xy)i + (a\xy + a2y2)).
Similarly, we have
proj(uX)a = (a • ux)ux = (~a\y + a2x){-yi + xj) = (1%\y2 - a2xy)i + (~a\xy + a2x2)).
This gives us
projua + proj(uX)a = {a\X2 + a2xy)i + {a.\xy + a2y2)) + (a\y2 - a2xy)i + {-a\xy + a2x2)j
= a\(x2 + y2)i + ci2(x2 + y2)j.
Since u is a unit vector, we have x2 + y2 - 1, so projua + proj(u±)a = aA + a2j = a, as desired.
(c) There are many ways to see why this equation holds. First, you might see that it is just the Pythagorean
Theorem! Let's see why. We saw above that projua = (a • u)u. Since u is a unit vector, the magnitude of
projua is a • u. Similarly, the magnitude of proj(uX)a is a • ux. In the diagram we used for our geometric
solution to part (b), triangle OXA is a right triangle with hypotenuse ||a|| and legs of lengths ||projua|| and
Pr°j(u-)a • Using our expressions above for ||projua|| and proj(u±)a , we have
l|a||2 = (a • u)2 + (a • ux)2
from the Pythagorean Theorem.
181
We also could have used the relationship we proved in part (b), rather than the Pythagorean Theorem.
Since u and are orthogonal, the projections of a onto these vectors are orthogonal:
projua • proj(uX)a = ((a • u)u) • ((a • ux)ux) = (a • u)(a • ux)(u • ux) = 0.
(There are no denominators in the expressions for the projections because u and ux are unit vectors.) We
therefore have
IMP = a • a = (projua + proju±a) • (projua + proju±a)
= (pr°jua) • (Proiua) + (prcva) • (p«va)+ 2(Proiua) • (Prova)

= ||pr°jua||2 + ||proju±a||2 + 2(0)
= (a • u)2 + (a • ux)2.
9.4.7 We have
u■v u•v u•v
(projvu) • (projuv) u•v v•u (u • v)3
rU
u lull2 llvll2
|u||2||y||2 Hull2 llvll2
sec2 0
(u • v)2 |u||2||v||2 cos2 0
Review Problems
9.29 We have = \J (-6)2 + 22 = 2 VlO [. and ||15i - 8j|| = y/152 + (~8)2 = \l7\.
9.30 A set of two vectors is linearly dependent if and only if one of the vectors is a multiple of the other vector.
Let 4i - 5j = t(-3i + kj) = -3fi + tk). Then 4 = —31 and -5 = tk, so t = -4/3, and k = —5/f - —5/(—4/3) — 15/4
9.31 First, we find the unit vector in the direction of 4i - 6j, which is
4i - 6j _ 4i-6j _ 4i-6j _ 4i — 6) _ 4i - 6j _ 2 Vl3i - 3 Vl3j

ll4i - 6jll ~~ ^42 + (_6)2 ~~ y/16 + 36 V52 2Vl3 _ 13
Then the vector v with magnitude 10 in the direction of 4i - 6j is given by
2 Vl3i - 3 Vl3j 20 Vl3i - 30 Vl3j

v = 10-
13 13
9.32 We have AB + BC + CA = (B - A) + (C - B) + (A - C) =
9.33 Let D = A + B and let O be the origin. We have C = \D, so C is the midpoint of D. Since OADB is a
parallelogram, diagonals OD and AB bisect each other. Therefore, the midpoint of OD is also the midpoint of AB,
which means C is the midpoint of AB.
9.34 Squaring the given inequality, we find that ||v|| + ||w|| > ||v + w|| is equivalent to
llvll2 + 2||v||||w|| + INI2 >||v + w||2.
182
Review Problems
Let 0 be the angle between v and w. We have ||v||2 = v • v, ||w||2 = w • w, and
llv + wll = (v + w)-(v + w) = v-(v + w) + w-(v + w) = v- v + v- w + w- v + w- w = v- v + 2v-w + w- w

= l|v||2 + ||w||2 + 2 ||v|| ||w|| cos 0.
Since cos 0 < 1, we have
||v + w||2 = ||v||2 + ||w||2 + 2 ||v|| ||w|| cos 0
<I|v||2 + ||w||2 + 2||v||||w||-(||v|| + ||w||)2.
Since the magnitude of any vector is nonnegative, we can take the square root of both ends to find that we have
||v + w|| < ||v|| + ||w||, as desired.
As we explained in the text, we can find the vector v + w by drawing w starting

from the head of v. The vector from the tail of v to the head of this w is v + w, as
shown. The expressions ||v + w||, ||v||, and ||w|| are the side lengths of this triangle.
So, the relationship we just proved is equivalent to the Triangle Inequality.
9.35
(a) (3i - 2j) • 8i = 3 • 8 + (-2) • 0 = 24 + 0 =
(b) = (-2) • (-4) + 5- 3 = 8 +15 =
9.36 In general, u • v = ||u||||v|| cos 0, where 0 is the angle formed by the vectors u and v. In particular, if u and v
have the same direction, then 0 = 0, so cos 0 = 1, and u • V = u
If u and v have opposite directions, then 0 = 7Z, so cos 0 = —1, and U V = -||u||||v||
9.37
(a) Let a = j, b = and c = (^j- Then a b and a • c = 0 gives us

• a\b\ + ^2 = 0 and a\C\ + CI2C2 = 0.
Since the vector a is nonzero, at least one of the components a\ and ci2 is nonzero. Without loss of generality,
assume that a\ is nonzero. Then from these equations, we have b\ = -02^2/01 and c\ = -02C2/01, so
f-a2b2c2/a-i\ _ f-a2b2c2/a1\
V b2c2 ) \ b2c2 )
If C2 = 0, then Ci = -02C2/01 = 0, so c = 0, in which case b and c are linearly dependent. If C2 is nonzero,

then the linear combination above shows that b and c are linearly dependent. Hence, in either case, b and
c are linearly dependent.
We can also reason geometrically. The equations a • b = a • c = 0 imply that a is perpendicular to both b
and c, so b and c are parallel. Hence, one vector is a multiple of the other vector, i.e. the set {b, c} is linearly
dependent.
(b) Yes. If c and b are linearly dependent and are both nonzero, then there is a nonzero constant k such that
c = kb. So, we have a c = a • (A:b) = ka ■ b = 0. However, we are given that a • c ^ 0, so we conclude that b
and c cannot be linearly dependent after all. Therefore, b and c are linearly independent.
9.38 Let v = Then u v = = 3x + 2y, so 3x + 2y = 9. Thus, the heads of all the vectors v form a
line
183
9.39 Since both sides of the given equation are nonnegative, squaring both sides gives the equivalent equation
||v + w||2 = ||v - w||2. But
||v + w||2 = (v + w) • (v + w) = v • (v + w) + w • (v + w) = v ■ v + v • w + w • v + w • w = v • v + 2v • w + w • w,
and
||v - w||2 = v • v + 2v • (-w) + (-w) • (-w) = v- v — 2v-w + w-w.
Hence,
||v + w||2 = ||v — w||2 <=> v-v + 2v-w + w-w = v-v — 2v-w + w-w <=> 4v-w = 0 v ■ w = 0.
Thus, ||v + w|| = ||v — w|| if and only if v and w are orthogonal.
9.40
(a) Let v and w be two vectors that have the same direction, and let u be the unit vector in the common
direction. Then both v and w are scalar multiples of u, because by definition, v = ||v||u and w = ||w||u.
Therefore, one of v and w is a multiple of the other.
(b) No. Let v be a nonzero vector, and let w = -v. Then w is a scalar multiple of v, but they do not have the
same direction, because they have opposite directions.
9.41 Yes. Let 0 be the angle between u and v. Then, we have u • v = ||u|| ||v||cos0. Since ||u|| and ||v|| are
nonnegative, the sign of u • v matches the sign of cos 0. Since cos 0 is positive for acute angles 0 and negative for
obtuse angles 0, then a positive dot product u • v means the angle between u and v is acute, and a negative dot
product means the angle between the vectors is obtuse.
9.42
v•u (0)(3) + (~3)(5) -iS /3\ /-45/34N

(a) projuv = rU -
U 34 \5J ~ V-75/34/
(V32 + 52)2
v•u (2) (5) + (—4)(7) -18 (5\ /-45/37\

(b) projuv = rU =
U 74 \7j ~ V-63/37/
(V52 + 72)2
9.43 The vector projvu must be a multiple of v. Letting projvu = kv, the equation projvu = u gives us u = tv.
Therefore, we have u + (—k)v = 0, which means that u and v are linearly dependent.
9.44 Algebraically, letting 0 be the angle between a and b, we have
a•b a • b| | COS 01
projba
iibiF
b
w \M
||b|| = ||a|| | cos 0|.
Since -1 < cos 0 < 1, we have | cos 01 < 1, so ||a|| | cos 0| < ||a||. Therefore, we have ||projba| < ||a||.
Geometrically, we showed in the text that if a is not in the same or opposite direction as b, then the arrow from
the origin representing a is the hypotenuse of a right triangle with the arrow representing projab as one of the
legs. The hypotenuse of any right triangle is longer than each leg of the triangle, so ||a|| > ||projba||. If a is in the
same or opposite direction as b, then projba = a, so ||a|| = ||projba||. Combining these, we have ||a|| > ||projba||.
Challenge Problems
9.45 If the set of vectors {u, v) is linearly dependent, then the set of vectors {u, v, w) is linearly dependent.
184
Challenge Problems
Otherwise, the set of vectors {u, v) is linearly independent. Hence, there exist constants a and b such that
au + bx = w. Rewriting this equation as au + bv + (-l)w = 0, we see that there exist constants Ci,C2,C3 such that
not all three are 0 and Ciu + C2V + C3W = 0. Therefore, the set of vectors ju, v, w} is linearly dependent.
9.46 If the vector ( has magnitude 6, then x2 + 4 = 36, then x2 = 32, or x - ±4V2. If the vector
2\
has magnitude 6, then 4 + y2 = 36, so y2 = 32, or y = ±4 y/2. Therefore, there are [I] such vectors:
y
2 W 2 N /4f),and(-4/
4V2/' V-4V2/'
Another way to think about this problem is to note that if fits the problem, then (x, y) is on the circle with
radius 6 centered at the origin, and either x = 2 or y - 2. Therefore each such (x, y) is the intersection of the circle
and either the line x = 2 or y = 2. Since each of these lines is less than 6 from the origin, each line intersects the
circle twice. Moreover, the lines meet at (2,2), which is not 6 units from the origin, so the lines meet the circle at
[~i] distinct points altogether.
Ml
9.47 Let u = /V M, and w = (^ ). Then u-v = «i-0 + 2- l=2, and v • w = 0 • zv\ + 1 • (-3) = -3. Also,
u w = uyw\ + 2 ■ (-3) = u\W\ - 6. Since u\ and w\ can be any real numbers, u • w can also be any any real number
9.48
(a) Geometrically speaking, let O be the origin, let X be (6, -2), so X = projab = 2a.
—^ -4 |
Let B be a point such that b = B and projab = 2a = X. Therefore, we must 1
.:j.
have BX _L 55. So, point B must be on the line through X perpendicular k B
u
to 55. Moreover, any point on this line can be B, including X. So, we r
5 / *
can take b = X = 2a = 6i - 2j. To find other possible values of b, we /n
note that 55 is the graph of y = -x/3. Rearranging this gives x + 3y = 0, proiQb
so the vector i + 3j is normal to 55. Letting n = i + 3j, we can then let

b = 2a + fn for any constant f
Checking, we see that if b = 2a + fn, we have
b • a = (2a + fn) • a - 2a • a + fn • a = 2 ||a||2 + f(0) = 2 ||a||
since n and a are orthogonal. Therefore,
b ■a 2 Hal
projab = a = 2a,
as desired.
(b) Continuing with the geometric set-up of part (a), we know that B is on the line through X that is perpen¬
dicular to 55. Since ||b|| = 10, we know that OB = 10. Therefore, B must also be on the circle with radius 10
centered at the origin. Since OX = \J62 + (-2)2 = V40 < 100, point X is inside this circle. This means that
the line that must contain X passes inside this circle, so this line intersects the circle at [2] points B such that
b = B satisfies both ||b|| = 10 and projab = 6i - 2j.
(c) If projjb = -j, then B is on the horizontal line through (0,-1). This line intersects the line through X
perpendicular to 55 at exactly [T] point, and the vector from the origin to this point is therefore the only
vector b that satisfies projjb = -j and projab = 6i - 2j.
185
9.49 Let U and V be points such that U = projur and V = projvr, and let O be the
origin. Li is in the same or opposite direction of u, and V is in the same or opposite
■"4
direction of v. Since u and v are linearly independent, so are U and V. Therefore,

neither U nor V is the origin, and U and V do not have the same direction or opposite
directions. This means that bl) and t)^ are two different lines. Let R be a point such
that r = R, and projur and projvr have the given values. Then R must be on the line
through U perpendicular to bll, as well as on the line through V perpendicular to
b\>. Since bll and OV are different lines through the same point, they have different
directions. Therefore, the two perpendiculars that pass through R cannot be the same
line, and cannot be parallel, which means they must intersect at exactly one point. So,
there is exactly one point R such that projur and projvr have the given values.
9.50 First, suppose projvu = 0. Then, we have v • u = 0, and from u = projvu +

projwu = 0 + projwu = projwu, we know that u = Aw for some nonzero constant k. Therefore, we have v • (Aw) = 0,
so v • w = 0.
Next suppose that both projections are nonzero. Let O be the origin and let U, V, and W be points such that
~+ ~+ —> —> ——+
U = u, V - projvu, and W = projwu. Since li = V + W, we know that OVUW is a parallelogram. Since OV is a
~+ —y
projection of U onto v, we know that VU _L OV, which means ZOVU = 90°. Since OVUW is a parallelogram, we
have /OWU = lOVU = 90° and /VUW = /.VOW = 180° - /OVU ~ 90°. So, OVUW is a rectangle, which means
OV _L OW, so V W = 0. We have V = k^v and W
■ few for some nonzero constants k\, k2, since V and W are
-
projections of U onto v and w, respectively. Substituting these into V -W = 0 gives v • w = 0.
9.51 Let (x, y) be a point on one of the angle bisectors. Any point on the bisector of an angle must be equidistant
from the sides of the angle. Therefore, (x, y) must be equidistant from 4x - 3y = 5 and x - 2y + 7 - 0. Applying
the formula for the distance between a point and line that we proved in the text, we have
|4x — 3y — 5| _ \x — 2y + 7|
y/'P'-t- (-3)2 y/l2 + (-2)2’
We have two cases to consider:
Case 1: The signs of 4x - 3y - 5 and x- 2y + 7 are the same. Then, we must have
4x - 3y - 5 _ x-2 y + 7
y42+(-3)2 “ yi2+(-2)2'
Multiplying both sides by 5 gives 4x - 3y - 5 = V5(x -2y + 7), and rearranging gives
(4 - V5)x + (-3 + 2 V% - 5 - 7 V5 = 0 .
Case 2: The signs of 4x - 3y - 5 and x-2 y + 7 are opposite. Then, we must have
4x-3y-5 _ f x-2y + 7
+ bp ” v A2+<-2)2
Multiplying both sides by 5 gives 4x - 3y - 5 = - x/5(x -2y + 7), and rearranging gives
(4 + V5)x + (-3 - 2 V% - 5 + 7 V5 = 0
186
CHAPTER
I - n
Matrices in Two Dimensions
10.1.1
(3 —4\ (3\ (3 • 3 + (-4) • (-5)\ 7 9 + 20 \ (29\

(a) \5 -2) V-57 ~ V5 • 3 + (-2) • (-5)7 \15 + 10) ~ V257
(-1 0W-lW(-l)-(-l) + 0-3\ 7 1 +0 \ (1)

(b)
U -2/ V 3 y - \4 • (-1) + (-2) • 3/ V(-4) + (-6)7 V-107
10.1.2 Let D = ( j11 j12 ). Since D = A - B, matrix D satisfies D + B = A, which means that we must have
V«21 “22/
7dll di2\ + 7^11 t»i2\ _ 7flll

1^21 ^22/ V&21 &22/ \a21 ^22/
Adding the two matrices on the left gives
7+ i’ll di2 + ^12^ _ 7aU fl12\

1^21 + ^21 ^22 + ^22/ \fl21 a22/
Therefore, we must have dn + bu = an, so dn = an - bn. Similarly, each entry in D is the difference of the
corresponding entries in A and B, so we have
. _ K _ n _ fau ~ bu an- bi2\

\fl21 ~ &21 a22 - ^22/
10.1.3
(a) We see that

-4\ (2\ 76-2 + (-4) • (-3)\ _ ,(12 + 12
-2) I-3J “ V3 • 2 + (-2) ■ (-3)/ 1V 6 + 6
and
-4\ 75>, 76 • 5 + (-4) • (-1)\ 730 + 4^
-2 ) \-v1 " \3 • 5 + (-2) • (-1)7 " \15 + 2y
187
CHAPTER 10. MATRICES IN TWO DIMENSIONS
(b) We showed in the text that a pair of vectors is linearly dependent if and only if one of the vectors is a
multiple of the other. Neither of the given vectors is a multiple of the other vector, so the set of vectors
2
is linearly independent.
-3
17 /24\
(c) The set of vectors is linearly dependent, because
12 ' V12)■
(d) For any vector we have
so A is always a scalar multiple of the vector ^ j. Therefore, for any vectors v and w, the set of vectors
{Av, Aw} is linearly dependent.
10.2.1
3 4 -1 0 3 • (-1) + 4 • (-5) 3-0 + 4-(-7) (-23 -28\

(a) -2 -6 -5 -7 (-2) • (-1) + (-6) • (-5) (-2) • 0 + (-6) • (-7) V 32 42 )
-4 -8 -i 2\ _ /(-4) •(-§) + (-8) • (-3) (-4) • 2 + (-8) • 3 (26 -32\

(b)
0 9 -3 3 0-K) + 9-(-3) 0-2 + 9-3 V-2 7 27 J
fli 0 bi 0
10.2.2 Let ^ o ^ J and ^ ^ ^ j be two diagonal matrices. Then their product is
fli 0 h 0 d\b\ 0
0 d2 0 b2 0 Q-ibi) ’
which is also diagonal
0 1
10.2.3 Let A = ,a b and F = Then
1 0
0 1 a b 0-fl + l- c 0-b + l- d
FA =
1 0 c d l-fl + 0- c l-b + 0-d
Thus, the matrix FA is the same as the matrix A when the rows are swapped.
Furthermore,
a b 0 1 a- 0 + b-l a-l + b- 0
AF =
c d 1 0 c-0 + rf-l c-l+rf-0
Thus, the matrix AF is the same as the matrix A when the columns are swapped.
0 1
10.2.4 The answer is no . For example, let A = B = ( ri ). Then
0 0
0 1 0 1 0-0 + 1-0 0-1 + 1-0 0 0

AB = = 0.
0 0 0 0 0-0 + 0-0 0-1 + 0-0 0 0
188
Section 10.3
10.2.5 Let T = ^. Then
f 4 \ _ (a b) ( 4 \ _ (4a - 3b\
T
V-3/ \c d) \-3j ~ \4c-3d) '
so from the given equation T , we have 4a -3b - -3 and 4c - 3d = 4. Also,
/-3a + b\
T
V-3c + d) '
so from the given equation T we have —3a + b 0 and -3c + d = 5. Hence,
/ 4 -3\ (a b\ (4 -3\ _ (4a - 3b -3a + b\ _ (-3 0\

V-3 1 ) ~ \c d) V-3 1 ) ~ V4c-3d -3c + d) ~ V 4 5J
4 -3
We also could have jumped straight to the answer by noting that the first column of T ( ^ j equals
4 -3
the product of T and the first column of ^ J, and the second column of the product is the product of T
and the second column of • We are giyen that the first product is ^ ^ j and the second is so
4 -3\ (-3 0
-3 1 4 5 '
10.2.6 If we multiply both sides of the equation A Q j = j by -2, then the left side becomes
-2A (2) = A (-2 (2)) = A (=4) .
and the right side becomes -2 f ^ ^j, so A f ^ ^. Thus, the two equations are the same.
10.3.1 Let A = g j). Then A (V-,) = („ J) (-2) = (o • 3 + 1 • (-2)) = (-2)' Hence' A is not necessarily
the identity matrix.
10.3.2 The reflection of the vector Q) over the x-axis is ( . Hence, we seek a matrix A = ^ such that
-y
a b\ (x x
c d) Vy -y
for all x and y. We see that ^ , so we can take a = 1, b = 0, c = 0, and d = -1, so the matrix
we seek is A =
189
10.3.3 Any linear function of the form /(v) = Av satisfies /(v + w) = /(v) + /(w) for all vectors v and w. But if
T
we take v = w = , then
/(v + w) =/ =/
and
l2 l2
/(v) + /(w) = /( ( 1 1 J + / 111
" U2 -
so /(v + w) A /(v) + /(w). We showed in the text that any function of the form g(v) = Av is a linear function, so
/(v) cannot be written in the form Av.
10.3.4 The matrix corresponding to a rotation of a + counterclockwise about the origin is
/cos(« + jS) - sin(« + jS)

\sin(a + p) cos (a +/3)
But such a rotation is also the composition of two counterclockwise rotations, with angles a and /?, respectively,
which corresponds to the product
/cosjS -sin/A /cos a -sin«\ _ /cos a cos jS - sin a sin - sin a cos fl - cos a sin
Vsin/1 cos/3 ) \sina cosn ) ~ Vsinacos^ + cosnsin/3 cos a cos - sin a sin jS
Therefore, cos(a + fi) = cos a cos jS - sin a sin and sin(a + /l) = sin a cos f> + cos a sin j8.
/I 0
10.3.5 As we found in Exercise 10.3.2, the matrix corresponding to reflection over the x-axis is A = -i)'an<^
the matrix corresponding to a 90° counterclockwise about the origin is B = ^ ^. Then the problem becomes
determining whether BA is equal to AB.
We see that
0 -1 1 0 _ (0 ■ 1 + (-1) -0 0 • 0 + (-1) • (-1)\ _ (0 1
BA =
1 0 0 -1 1 ■ 1 + 0• 0 1-0 + 0- (-1) 1 0
and
1 0 \ /O -l\_/l-0 + 0-l l-(-l) + 0-0 \ _ / 0 -1
AB =
o -1) \i o ) ~ vo - o + (-1) • i o • (-1) + (-1) • oj “ V-i o
These two matrices are not the same, so the two operations of reflecting over the x-axis and rotating 90° counter¬
clockwise about the origin do not commute, i.e. the order in which you perform them makes a difference.
10.4.1 First, we rewrite the given equations as
6x + 4y = 7,
2x + 5y = 9.
By Cramer's rule, we have
7 4 6 7
9 5 _ 7•5-4•9 _ 1 2 9 _ 6•9-7•2 40 20
/ and y= 6
6 4 ~ 6•5-4•2 “ 22 4 ” 6 • 5 - 4 ■ 2 ' 22 " 11
2 5 2 5
190
Section 10.5
10.4.2
-3 -5
(a) = (-3) • 18 - (-5) • 6 = -54 - (-30) = -24
6 18
(b) To make the determinant easier to evaluate, we take a factor of 800 from the first row and a factor of 1/3
from the second row:
2400 3200 3 4 800 3 4 800, s 800 . , 5600

= 800 = -(3-31-4-25) = — -(-7) =
25/3 31/3 25/3 31/3 3 25 31 3
10.4.3 For all matrices A and B, det(AB) = det(A) det(B) = det(B) det(A) = det(BA).
!0.4.4 LetA=(“ *),»B=(j ‘tfalrnd
a b + ka
det(B) = = a(d + kc) - (b + ka)c - ad + kac -be- kac = ad-be = det(A).
c d + kc
10.4.5
1 0\ , _ _ (0 0\ . _ _ fl 0\ (0 0 1 0
(a) Let A = Then
0 o j and B V0 iJ'soA + B (n oj + VO 1 0 1
1 0 0 0
det(A) -- = 1 • 0 - 0 • 0 = 0, det(B) = = 01-0-0 = 0,
0 0 0 1
and
1 0
det(A + B) = = l- l- 0- 0 = l.
0 1
So the conditions det(A) = 0 and det(B) = 0 are satisfied, which means det(A) + det(B) = 0 + 0 = 0, but
det(A + B) = 1, so det(A) + det(B) and det(A + B) are not necessarily equal.
(b) Let A = Q °)andB=Q JY so A + B = (l !T) + (* = (l ?V Then

0 0 0 1 0 1
I 0 1 0
det(A) = l- 0- 0- 0 = 0, det(B) = = 11-0-0 = 1,
0 0 0 1
and
2 0
det(A + B) 2 • 1 - 0 • 0 = 2.
0 1
So the conditions det(A) = 0 and det(B) t 0 are satisfied, but det(A) + det(B) = 0 + 1 = 1 and det(A + B) = 2,
so det(A) + det(B) and det(A + B) are not necessarily equal.
/ cos Q - sin 0\
10.5.1 LetR = . The determinant of R is given by
Vsin 6 COS0 )
cos 9 - sin 6
- cos 9 • cos 9 - (- sin 9) ■ sin 9 = cos2 9 + sin2 0 = 1.
sin 6 cos 9
191
As discussed in the text, multiplying a vector by this matrix V

rotates the vector. Rotation does not affect area (or signed area),
so the parallelogram with adjacent sides Ri and Rj is a rotation of
the parallelogram with adjacent sides i and j. This latter parallel¬ (- sin 9, cos
ogram is a square with side length one, and rotation doesn't affect
signed area, so the parallelogram with adjacent sides Ri and Rj
also has area 1. Therefore, the matrix has determinant 1.
Another way to look at this is that Ri = (cos 6)i + (sin 6)j and
Rj = (- sin @)i + (cos 9)j. The point (- sin 9, cos 9) is obtained by
rotating the point (cos 9, sin 9) by 90° counterclockwise about the
origin, so Rj is a 90° counterclockwise rotation about the origin. Moreover, the magnitudes of Ri and Rj are both
equal to 1. Hence, the parallelogram with adjacent sides Ri and Rj is a square with side length 1, and therefore
has area 1.
10.5.2 We know how to find the area of a triangle given its vertices, so we split the
quadrilateral into two triangles, as shown. As described in the text area of triangle ABD is
given by
1 1 - (-2) 8 - (-2)\
det ||3 • (-2) - (-8) -10| = ||74| = 37,
2 -5-3 1-3 J
and the area of triangle BCD is given by
1 ( 7-1 8-1 \ 1 1 29
det -|6-6-1.7| =-|29| = y.
2 \—4 — (—5) 1 -(-5)J
so the area of quadrilateral ABCD is 37 + y =
10.5.3 Let v and w be the column vectors of A, so that v is the column we

multiply by k. Let A' be the resulting matrix when v is multiplied by k.
Suppose first that k is positive. Geometrically, det(A) is the area of the parallel¬
ogram formed by v and w, and det(A') is the area of the parallelogram formed by
kv and w. In the diagram at right, OVSW is the first parallelogram and OV'S'W
is the second. The area of a parallelogram equals the length of a side times the
height to this side from a point on the opposite side. The height from W to side
OV (extended if necessary) equals the height from W to OV. Therefore, the ratio
of the area of OV'S'W to the area of OVSW equals OV'/OV, which equals k.
192
Section 10.6
If k is negative, the situation is essentially the same, except now, as shown y

in the diagram at right, the orientation of parallelogram OV'S'W is opposite
the orientation of OVSW. Again, we have OV'/OV = \k\, and the ratio of the
signed areas of the parallelogram is k.
Since the area of the parallelogram with sides kv and w is always k times
the area of the matrix with sides v and w, the determinant of the matrix with >
r^^7 w
columns kv and w is k times the determinant of the matrix with columns v
and w.
// 0 / X
kv/
10.5.4 Let the vertices of T be the lattice points (x\, y\), (X2, yi), and (X3, y3),
so all the coordinates are integers. Then the area of triangle T is
V'
det(X2~Xl X3"Xl
V1/2 - y\ ys- y\
When we double this quantity, we get
(X2-X\ x3- x1
det = $x2 - xi)(y3 - y$ - (*3 - xi)(i/2 - y\)\,
Vyi -y 1 3/3 - yi
which is clearly an integer.
10.5.5 We know that det(A) is the signed area of the parallelogram formed by the column vectors of A. Since
multiplying A on the right by F swaps the columns of A, the parallelogram formed by the column vectors of AF
is the same as the original parallelogram, except that the orientation has been reversed, so det(AF) = - det(A).
But det(AF) = det(A) det(F), so det(A) det(F) = - det(A). For this to hold for all matrices A, det(F) must be
equal to
10.6.1 Since I -1 — I, the inverse of I is [T|.

10.6.2
/ i 1\
6 8 r1 1 /4 -8 1 10 5 |
(a) -2 4, ) 6.4 - 8 • (-2) V2 6 40
\ i 3 /
\20 20 /
2 -1 1 1 0.5 1 1 (0.5 10\

(5
(b)
-0.9 0.5 2 • 0.5 - (-1) • (-0.9) V0.9 2) 0.1 V0.9 \9 20 )
40 30\ = (-40)
/ .m • (-6)
/ n -8
o • 30
on = n j-i
_ 0, the—(
matrix ( g^0 30 \ does not have an inverse.
(c) Since det
10.6.3
If A is nonsingular, then by definition, det(A) is nonzero, so A is invertible. Hence, we can multiply both
(a)
sides of the equation Aw = v by A-1 to get A"1 Aw = A_1v, which simplifies as w = A~!v |. So, | yes |, there
must be a vector w such that Aw = v.
No, even if A is singular, it is still possible that there is a vector w such that Aw = v. For example, if
(b)
^ 0 j ancj v _ then any vector w satisfies the equation Aw = v. If A is singular, all we can
conclude is that the equation Aw = v does not necessarily have a solution w.
193
10.6.4 We have that det(M) det(M x) = det(MM 1) = det(I) = 1. However, since all the entries of M and M 1 are
integers, det(M) and det(M_1) must be integers as well. The only integers that have 1 as a multiple are [T] and
If M = I, then M 1 = I and det(M) = 1, and if M = , then
v-mi
and det(M) = 1 • (-1) -00 = -1, so both 1 and -1 are possible values of det(M).
:h: ",)•
10.6.5 Let A = B = I = (J °),so A + B = (J J) + (J = (jj °)- A_1 = B_1 = I_1 = l' so
B1 + A-1 = 21 = (q 2) / and
(A + B)-1 =
2
0
0
2 2 2
1 2
0 0 \0
0
2
1 (2
4 V0
0
2
= ii ?
• - •
0 2
Hence, (A + B) 1 need not equal B 1 + A L (In fact, it's possible that A + B is not even invertible! For example,
consider the case A = I and B = —I. Then, both A and B are invertible, but A + B is not!)
10.6.6 For any complex number z, if we multiply z by a + bi, and then divide the product by a + bi, the result is z.
x
Correspondingly, if we multiply a vector ( I by the matrix L 1 and then multiply the result by the matrix
y
M that corresponds to dividing by a + bi, we should get (~ J back in the end. In other words, multiplying by M
"undoes" multiplication by f\b a )' means that the desired matrix is the inverse of . We have
fa -b\ 1 _ 1 / a b\ _ 1 fa -b\ / a b \
]
a2+b2
aY
\b a) a ■ a - (-b) ■ b \~b a) a2 + b2\b a) \a2+b2 a2+b2 )
(Note that the determinant of our matrix for multiplying by a + bi is a2 + b2 -\a + bi\2, which is nonzero if a + bi ± 0,
so this matrix is indeed invertible.)
Review Problems
10.42 We have
6 6c be a + 6c be + 4
+c
-3 -1 -3c -c —3c + 7 —c + 2
so a, b, and c must satisfy the equations a + 6c = 32, be + 4 = 9, -3c + 7 - -8, and -c + 2 = -3.
From the equation -c + 2 = -3, we have c = 5. This value of c also satisfies the equation -3c + 7 = -8. Then
from the equation a + 6c = 32, we have a = 32 - 6c = 32 - 6 • 5 = 2, and from the equation be + 4 = 9, we have
5b + 4 = 9, so b = (9 - 4)/5 = 1. To summarize, all four equations are satisfied by a = 2, b = 1, and c = 5
10.43 We have
-3 2 • (-3) + 3-5 2 - (—1) + 3-2

MN =
-7 5 (-7) • (-3) + 0-5 (-7) • (-1) + 0-2
194
Review Problems
and
NM =
-3 2 (-3) • 2 + (-1) • (-7) (-3) • 3 + (-1) ■ 0
5 -7 5 • 2 + 2 • (-7) 5•3+2•0
10.44 Solution 1: Use the determinant. Let the matrix be ^ ^ j. We know that the rows of the matrix are linearly
dependent if and only if the determinant is 0, i.e. ad-be = 0.
Now consider the matrix L A. The rows of this matrix are linearly dependent if and only if its determinant
is 0, i.e. ad - be = 0. Furthermore, the rows of this matrix are the same as the columns of our original matrix.
Hence, the rows of a matrix are linearly dependent if and only if the columns of the matrix are linearly dependent.
Solution 2: The longer way. Let the matrix be ■ If the rows are linearly dependent and no entries are
0, then a - ct and b = dt for some constant t. Then, we have " = so | = Letting these ratios equal k, we
have a = bk and c = dk, so the first column is k times the second, which means the column vectors are linearly
dependent. If one entry is 0, then we have two cases to consider: either the other entry in the row is 0 or the
corresponding entry in the other row is 0, because the rows are linearly dependent. A little more casework shows
that in both cases one column is a multiple of the other. (The casework caused by 0 makes this the much longer
way.)
Similarly, if the columns are linearly dependent and no entries are 0, then a = bk and c - dk for some
constant k, from which we find | = Rearranging this gives “ = |, so one row is a multiple of the other, meaning
the rows are linearly dependent. As before, if one entry is 0, we can work through some more simple cases to
show that the rows are linearly dependent.
10.45 Suppose that such a matrix F = y exists. Then

w
sc ?)=(?:
But
fi=f=(* q,
Vz w)
so x = 0, y = 1, z = 1, and w = 0. Therefore, if such an F exists, it must be F = ^ ^. But for an arbitrary matrix
A=(“ bX we have
0 1 \ fa b\ _ [0-a+ l-c 0-b + l-d\ _ fc d

FA =
V1 0) \c d) Vl-fl + 0-c l-b + 0-d) \a by
but (C f) is not the matrix formed when the columns of A are swapped. Hence, no such matrix F exists.
\a bj
10.46 From the given information, we have
9 -28
A(3v - 4w) = A(3v) - A(4w) = 3Av - 4Aw = 3 -4f ^
-6 16
195
10.47
(a) LetAi=(J Then
1 0 1 0 1+0 1 -0 + 0-0
A? = 0 0 0 0 1+0 o-o + o-o
Also, let A2 = f q ^ ). Then
0 0 0-0 + 0 0 0-0 + 0-1

A2 - (° 0
A2 “ IQ 1 0 1 0-0 + 1 0 0-0 + 1-1
0 1
(b) Let A = . Then
1 0
0 1 0 1 0-0 + 1-1 0-1 + 1-0

A2 = = 1,
1 0 1 0 1 -0 + 0-1 1-1 + 0-0
so A2 ± A and A3 = A2 • A = IA = A.
There are many other matrices that satisfy this part. Any matrix such that A2 = I and A ^ I will do. For
example, a matrix A will satisfy A2 = I if Av is a reflection of v over some line for any vector v. This is
because multiplying Av by A (to give A2v) just reflects Av back over the line, which results in v.
10.48
(a) The reflection of over the y-axis is so we seek a matrix A such that
for all x, y. We can take a - -1, b = 0, c = 0, and d = 1, so we can take A = (-1 °"l
vo V
a b
(b) The reflection of ( ~ J over the graph of x = y is , so we seek a matrix B = ( “ "A such that
for all x, y. We can take a = 0, b - l, c - l, and d - 0, so we can take B
10.49 A clockwise rotation by 0 is equivalent to a counterclockwise rotation by -0. We can therefore substitute
-0 as the angle into the rotation matrix we found in the text, to find
( cos(-d) - sin(-0)\ / COS0 sin 0\

Vsin(-0) cos(-0) / sin 0 cos 0/
10.50 We see that
/ cos 0 - sin 0 \ f COS 0 - sin 9\ f cos2 0 - sin2 0 -2 sin 6 cos 6 \

Vsin 0 COS0 J
Vsin 0 cos 0 J \ 2 sin 0 cos 6 cos2 6 - sin2 6 J '
196
Review Problems
But by the double angle formulas for cosine and sine, cos 20 = cos2 6 - sin2 6 and sin 29 = 2 sin 9 cos 9, so
j2 _ (cos 20 - sin 20N

V sin 20 cos 20
Thus, T2 corresponds to a rotation counterclockwise about the origin by an angle of 20.
We also have
T3_tt2=^COS^ - sin 0\ /cos20 -sin20V /'cos 0 cos 20 - sin 0 sin 20 - cos 0 sin 20 - sin 0 cos 20
,sin0 cos0 / \sin20 cos 20 J Vsin 0 cos 20 + cos 0 sin 20 - sin 0 sin 20 + cos 0 cos 20
(cos(0 + 20) - sin(0 + 20)\ _ /cos30 -sin30
Vsin(0 + 20) cos(0 + 20) ) ~ Vsin30 cos30
10.51 We see that

cos 45° - sin 45°
M“(v^ vf)=2(| f)=2 sin 45° cos 45°
/cos 45° -sin45°\ , ._0 , , . , ...

The matrix V sin 45° cos 45° J corresPonc*s to a 45 counterclockwise rotation about the ongm, so
10
cos 45° - sin 45° /cos(10 • 45°) - sin(10 • 45°) cos 450° -sin 450° 0 -1
sin 45° cos 45° \sin(10 • 45°) cos(10 • 45°) sin 450° cos 450° 1 0
Therefore,
10
cos 45° -sin 45° cos 45° ( 0 -1024V
M10 = ( 2 = 2 10 Sto45V” = 1024 0 ‘J
sin 45° cos 45° sin 45° cos 45° V1024 0 )
10.52 To recover point P, we can take the point (-2 - V3,1-2 \/3), rotate 60° clockwise about the origin, and
then reflect over the y-axis.
Rotating 60° clockwise about the origin, we get
(cos(-60°) -sin(-60°)\ f-2- V3\ / \ (-2- V3" i • (-2 - y/3) + f - (1 - 2 a/3) -4

\sin(-60°) cos(-60°) J 1 1 -2 V3 J [-4 11 1-2V3 2
2 ,(--^r) • (-2 - V3)+i-(l-2V3),
and the reflection of the point (-4,2) over the y-axis is (4,2). Therefore, P = (4,2)
10.53 For any linear function /, we have /(0) = /(0 • 0) = 0 • /(0) = 0.
10.54
7 -4
(a) = 7 • 7 - (-4) • (-2) = 41
-2 7
0.27 -175
(b) = 0.27 • 95 - (-175) • (-0.15) = -0.6
-0.15 95
10.55
(a) Let A and B be two singular matrices, so det(A) = det(B) = 0. Then det(AB) = det(A) det(B) = 0, so the
matrix AB is singular. Hence, the product of two singular matrices is also singular.
(b) Let A and B be two nonsingular matrices, so det(A) + 0 and det(B) £ 0. Then det(AB) = det(A) det(B) ± 0,
so the matrix AB is nonsingular. Hence, the product of two nonsingular matrices is also nonsingular.
197
(c) Let A be a singular matrix and let B be a nonsingular matrix, so det(A) = 0 and det(B) + 0. Then
det(AB) = det(A) det(B) = 0, so the matrix AB is singular. Hence, the product of a singular matrix and a
nonsingular matrix is always singular.
10.56 Let A = I = Q ^. Then C = Q ^, so det(C) = det Q ^ = 1 - jk. However, det(A) = det(I) = 1, so
det(A) and det(C) are not necessarily equal.
10.57 Recall that the determinant of a matrix is the signed area of the parallelogram formed by the column
vectors of the matrix. Swapping the columns of the matrix reverses the orientation of this parallelogram, and
reversing the orientation of the parallelogram changes the sign of the signed area.
10.58
3\ 1 _ 1 f8 -3\
(a)
8) ~ (-2) ■ 8 - 3 • (-2) V2 -2)
-3\_1 _ 1 fO 3\
(b)
0 ) ~ 0 • 0 - (-3) • (-2) \2 0)
(c) Since det = 6 • (-4) - 12 • (-2) - 0, the matrix does not have an inverse.
10.59 Since A 1A = I, the equation A 1 = A tells us that A2 = I, sodet(A2) = det(I) = 1. Butdet(A2) = [det(A)]2 =
1, so det A = 1 or -1.
If A = I, then A-1 = I-1 = I = A, and det(A) = det(I) = 1. If A = Q then
2_/l 0 W1 OV/l-l + O-O 1 • 0 + 0 • (-1) \ _ A 0\

V0 -1/ V0 -1/ V0-1+ (-!)• 0 0 • 0 + (-1) • (-1)/ _ vo l)~1.
so A 1 = A, and det(A) = det (* = 1 • (-1) - 0 • 0 = -1. Hence, values of [l] and | -11 are both possible.
10.60
fa + x b + y\
(a) LetA=(“ J) and B = (* .^),soA + B “
c
b) + \z
a)
(x wj
y) Vc + z d + w)
, and
At =
so
fa c\ fx z\ fa + x c + zV
At + Bt
\b )
d + \y w) \b + y d + w) '
and
(A + B)t =
fa + x b + y) fa + x c + z\
Vc + z d + w) \b + y d + w) '
Hence, (A + B)r = AT + Br.
(b) Let A = ^ Then det(AT) = det C^j = ad - be = det(A). Furthermore, A is invertible if and only if
det(A) ^ 0. Therefore, if A is invertible, then Ar is invertible.
198
Challenge Problems
(c) Let A = H ^ ), where A T - I a c ' is invertible. Then

b d,
a c d —c
(AT)"1 =
b d ad - be \~b a
Also,
-l
a b d -b
A"1 =
c d ad - be \-c a
so
d -c
(A"1)7 =
ad - be \-b a
Hence, (AT)-1 = (A_1)T.

-1 -6
(d) Given R = ^ and S = /we have rT ~ (3 4”) anci S? = , so
2 V (-1 5) / 2 • (-1) + 3 • (-6) 2-5+ 3-3 \ /-20 19\

RS = ^ V-23 7)
V-6 3j - \(—1) • (_1) + 4 • (-6) (-1) -5 + 4-3/
20 19\T (-20 -23\

(RS)7
23 7J { 19 7 )
2 -1\ (-1 -6 2 • (-1) + (-1) • 5 2 • (-6) + (-1) ■ 3^ _ (~7

-15\
RrSr = 3 • (-1) + 4-5 3-(-6)+ 4-3 \17 -6 )
3 4 ) \ 5 3
-1 -6\ (2 -1 (-1) • 2 + (-6) • 3 (-1) • (-1) + (-6) • 4^ _ (—20 -23\

strt = 5-2 + 3-3 5 • (-1) + 3-4 V 19 7 )
Note that (RS)T = STRT. We claim this is true in general.
LetR=(“ 2)andS=(* */,» °
a b\ (x y\ _ (ax + bz
RS =
c d) \z w) \cx + dz cy + dwj'
yandsT=w
x z\ fa c\ _ ( ax + bz cx + dz\
strt = y w) \b d)~\ay + bw cy + dwj'
Hence, (RS)r = STRT for all matrices R and S.
Challenge Problems
1 0 2 0 2 0
10.61 Let A = ( Q 2J then B = ^ ~ j. Then AB = BA = ^ ~ j. To find examples of A and B where none of
the entries are equal to 0, we can use rotation matrices. Let
/cos60° -sin 60°^ _ .^T

/cos30° -sin30°^ _ (^ 4 B =
A = and V3
V sin 30° cos 30° Vl V sin 60° cos 60°
2 2 2 -
199
cos 90° - sin 90° 0 -1

Then AB = BA -
sin 90° cos 90° 1 0
10.62 If A6 = A, then det(A6) = det(A). But det(A6) = [det(A)]6, so [det(A)]6 = det(A).
Let x = det(A), so x6 = x. Then x6 - x = 0, so x(x5 - 1) = 0. Therefore, either x = 0 or x is a fifth root of unity.

The only fifth root of unity that is real is 1, so if all the entries of A are real, then det(A) must be equal to 0 or 1
0 0
If A = , then A6 = A, and det(A) = 0, and if A = I, then A6 = A = I, and det(A) = 1. Hence, both 0 and
0 0
1 are possible values of det(A).
co 0
If the entries of A can be nonreal, then in addition, det(A) can be any fifth root of unity. Let A = ( ^ ^
where co is a fifth root of unity. Then

_ co6 0 co 0
= A,
0 1 0 1
and det(A) = co.
10.63 If A4 = A, then A3 - I. So, we seek a matrix A such that multiplying any vector by A three times leaves
the vector unchanged. Rotation by 120° is a transformation that leaves a vector unchanged if applied to the vector
three times. We expect that the rotation matrix for a 120° counterclockwise rotation will fit the problem. Let's
check. We let
fcos 120° -sinl20°^ _ (~\2
sin 120° cos 120° V3
2
Then
/cos 480° - sin 480°\ /cos 120° -sin 120° \
A4 = = A,
V sin 480° cos 480° ) ■ V sin 120° cos 120° )
but
cos 240° - sin 240° 1 VT
2 t A.
sin 240° cos 240° VI
2 ~ 2'
So, we have A4 = A, but A2 =£ A, as desired.
04 U2 b\ h
10.64 Let A = and B = . Then
03 «4 h h
04 + b\ 0-2 + b2
det(A + B) = det
03 “t b2 #4 + &4
- («! + frl)(Q4 + ^4) — 0?2 + f’2)(fl3 + ^3)
— 0404 + 04/74 + 04 b\ + £74/74 — 0203 — 02^3 — 03^2 — b2b3,
and
det(A - B) = det ("> _ £ £_ = («, - h)(a4 - bt) - (a2 - b2)(a3 - b3)
= 0^04 — 04/74 — 04/74 + £74/74 — 0203 + 02^3 + 03^2 — b2b^,

SO
det(A + B) + det(A - B) = (0404 + 04/74 + 04/74 + £74/74 - 0203 - 02/73 - 03/72 - b2b^)
+ (0404 — 04/74 — 04/74 + £74/74 — 0203 + 02^3 T 03/72 — b2b^)
= laiai + 2/74/74 - 20203 ~ 2/72Z>3 = 2(0404 ~ 0203) + 2(/>4/74 - /72/73) = 2[det(A) + det(B)].
200
Challenge Problems
10.65 If we compute the determinant, we get
(x - *i)(y - y2) - (x - x2)(y - y\) = 0.
When we expand the left side, the xy terms cancel and we are left with an equation of the form ax + by + c = 0,
where a, b, and c are constants. The graph of such an equation is a line. Furthermore, if (x, y) = (xi, yi), then
X - Xi y-yi 0 0
X - X2 y-yi X\ - X2 yi-yi
and if (x, y) = (x2, y2), then

X- X\ y-yi X2 ~ Xi y2-y\
x - x2 y-yi 0 0
Thus, both points (xi,yi) and (x2,y2) satisfy this linear equation. Hence, it represents the equation of the line
passing through both points.
10.66
(a) We prove the result by induction. For n = 1, we have
A =
so the result holds for n = 1. Assume that the result holds for some positive integer n-k, so
k _ (Fk +1 Fk \
V Fk Fk-iJ
Then
(1 1\ (Fk+1 Fk\ (Fk+i + Fk Fk + Fk-i\ (Fk+2 Fk+i\
Afc+1 = A • Ak -
U o) V Fk Fk-iJ “ V Fk+1 Fk ) \Fk+1 Fk )
Hence, the result holds for n = k + 1, so by induction, it holds for all positive integers n.
(F i F \
F+n F„-.J'We§et
det(A'!) = det p"-^) = Fn+iFn-i ~ F„-
But
1 1
det(A”) = [det(A)]” = det = (-1)",
1 0
so F„+iF„_i -F2n = (-l)n.
10.67 We see that , . N , Q • a\

cos d - sin 0\ _ /x cos Q — y sm @\ _ /3\
sin0 cos0 ) \y) ~ \xsm0 + ycosdj \2J '
In other words, if the vector Q) is rotated by an angle of 0 counterclockwise about the origin, then the vector
^ ^ obtained. The only such vectors are the vectors such that (x, y) is on the same circle, centered at the
2/ y
origin, as (3,2). Hence, there exists such a 6 if and only if|xz + y2 - 13
201
10.68
(a) Let t be a line through the origin and let /(v) be the reflection of v over t. To show that / is linear, we must
show that /(v + w) = /(v) + /(w) and that f(kv) = kv. The diagram at left below exhibits the former and the
diagram at right exhibits the latter.
In both diagrams, b is the graph of a line through the origin. In the diagram at left, parallelogram OVSW
is constructed, and OV'S'W is the reflection of OVSW over £. Therefore f(V + W) = f(S) = S' and
f(V) + f(W) = V' + W' = S'. In the diagram at right, T - kV, and V and T are the reflections of V and T,
respectively, over l, so f(T) = V. By SAS Similarity, we have A VOW ~ A TOT', so OT'/OV' = OT/OV = k,
which gives T' — kV'. Therefore, we have f(kV) = f(T) = T' = kV'. We can go through essentially the same
steps if k < 0. We can therefore conclude that / is linear.
(b) Since the reflection is a linear transformation, we can represent the action of the transformation on any
vector as a product of the matrix and the vector. So, we just have to find the matrix A such that Av is the
reflection of v over y = 2x for any vector v. We know the results of two reflections, and can use these to
find A. Specifically, the vector is in the direction of the line, so it is its own reflection. Rewriting the
2
equation as 2x - y = 0, we see that | is normal to the line, so it is the opposite of its reflection. Therefore,
we have
/i\ /i\ 2 -2
and
-1 1
Letting A = , we have
u + 2& 2a - b -2
and
c + 2d 2c -d 1
From the system a + 2b = 1, 2a - b = -2, we have a = b = From the system c + 2d = 2, 2c - d = 1, we

/_ 3 4'
have c = I, d = 1 so A =
To find the reflection of (4, -6), we note that

_3 36
7
5
so the desired point is HM)

(c) If / is linear, we must have /(0 + 0) = /(0) + /(0), which gives /(0) = 2/(0), so /(0) = 0. However, the
reflection of (0,0) over x + y - 6 is not the origin, so / cannot be linear.
202
Challenge Problems
10.69
(a) Let v =
y)' Then Av - ( 2 -l) (y) = ( 2x-yy)' and Av = (ly)' S0 Av = Av §iveS the system
of equations -2x + 6y = Ax and 2x - y - Ay. Then 6y = (A + 2)x, so y = (A + 2)x/6. From the equation
2 x-y - Ay, we have
2x = (A + l)y = (A + 1) • Â + = (A + 1KA + 2)xj

6 6
If x = 0, then y = (A + 2)x/6 = 0. But we seek nonzero vectors v, so we may divide both sides by x in the
equation above to get
2 = -A-+ 1^(A + 2) =* (A + 1)(A + 2) = 12.

6
This simplifies as A~ + 3A - 10 = 0, so (A - 2)(A + 5) = 0, which gives A = 2 or A = -5 as the only possible
values of A. So, we seek vectors v = ^ J such that either Av = 2v or Av = —5v.
For Av = 2v, we have —2x + 6y - 2x and 2 x-y - 2 y, both of which give 2x = 3 y. So, we can take x = 1,
and y - 4 • 1/6 = 2/3, which gives v = (2/3) • Any multiple of this vector also satisfies Av = 2v
For Av - -5v, we have —2x + 6 y - -5x and 2 x-y = -5 y, both of which give x - -2 y. we can take x = 1,
and y = -1/2, so v = * Any multiple of this vector also satisfies Av = -5v.
(b) For A = 2, we have
-2 6 1 0
det(A - AI) = det -2 = det = (-4) - (-3) -6-2 = 0,
2 -1 0 1 2 -3
and for A = -5, we have
det(A - AI) = det

-2
2 ; = det ( 2 f)=3-4-6-2 = 0.
If Av = Av, then we have (A - AI)v = Av - Av = 0. Furthermore, if det(A - AI) ^ 0, then A - AI is an

invertible matrix. Multiplying both sides of (A - AI)v = 0 by this inverse gives us v = 0. So, we cannot
have a nonzero solution v to the equation Av = Av if det(A - AI) =£ 0. Therefore, if there exists a nonzero
vector v that satisfies (A - AI)v = 0, then we must have det(A - AI) = 0.
10.70 First, we give an algebraic proof. Let A = I ^ j, so
a b
det(A) = = ad - be.
c d
We are given that the columns of A are orthogonal, so ^ = O' which means ab + cd = 0. The product of the
norms of the columns is Vfl2 + c2<W + d2. To make this quantity easier to work with, we square it, to get
(a2 + c2)(b2 + d2) = a2b2 + a2d2 + b2c2 + c2d2 = (a2b2 + 2abed + c2d2) + (a2d2 - 2abcd + b2c2)
- (ab + cd)2 + (ad - be)2 = 0 + (ad - be)2 = (ad - be)2,
so Vfl2 + c2 Vb2 + d2 = I ad - bc\ = \ det(A)|.
203
Now we give a geometric proof. We know that | det(A)| is the area of the parallelogram
formed by the vectors v and w = But we are given that these vectors are
orthogonal, so the parallelogram is a rectangle. This means its area is simply the product of x
the lengths of consecutive sides, which is equal to the product of the norms of v and w.
10.71
(a) In the text we showed that the area of a triangle with vertices (xx, yx), (x'2, y2), and (X3, y3), in counterclockwise
order, is
x2 - Xi x3 - Xi 1 1
= ^((*2 - xiXys - yi) - (y2 - yi)(*3 - *i)) = ^*\yi + *zy3 + *3yi - *2yi - - *iy3).
yi - yi y3 - yi
which is exactly the expression given by the Shoelace Theorem.
(b) We split the quadrilateral into a triangle with vertices (xi, y\), (x2, y2), and (X3,3/3) and a triangle with vertices
(X3,1/3), (X4, t/4), and (xi, yi). We then apply part (a) to both triangles, which tells us that the area of the first
triangle is
1-(xiy2 + x2y3 + x3yi - x2yx - x3y2 - xxy3)
and the second is

(x3y4 + x4yi + xxy3 - x4y3 - xxy4 - x3yx).
Adding these gives
-(xxy2 + x2y3 + x3y4 + X4yx - x2yx - X3y2 - X4y3 - xxy4).
which is again exactly the expression given by the Shoelace Theorem.

(c) We apply induction. We assume the Shoelace Theorem gives the area for a polygon with k sides, where k > 3.
We let the vertices, in counterclockwise order, of a polygon with k+1 sides be (xx, yx), (x2, y2),..., (xx+1, y^+i).
We then break this polygon into the k-sided polygon with vertices (xx, yx), (x2, y2),..., (xk, y^) and the triangle
with vertices (xx, yk), (xx+x, y^+i), (xx, yx). Applying the inductive assumption, the area of the k-sided polygon
is
1
-(xxy2 + x2y3 + • • • + xjt-iyk + xkyx - x2yx - x3y2-x^y^-i - xxyk)
and the area of the triangle is
^{xkyk+\ + Xk+iyi +XXyk- x,t+iyk - xxyk+1 - xkyx).
Adding these gives the total area of the k + 1-sided polygon as
-(xxy2 + x2y3 + • • • + xkyk+1 + xk+1yt - x2yx - x3y2-xk+lyk - xxyM),
and the induction is complete.
10.72
rA and
a s = (s1 , and take B = rx si
(a) Let r = . Then
ri S2 r2 s2
n sx\ (\
Bi r.
r2 s2) VO
and
r\ sx\ (0
Bj - s.
r2 s2
204
Challenge Problems
(b) The signed area of parallelogram with r and s as sides is det(B).

(c) Since Ar = A(Bi) = (AB)i and As = A(Bj) = (AB)j, the signed area of the parallelogram with Ar and As as
sides is det(AB).
(d) Since det(AB) = det(A) det(B), the signed area of the parallelogram with Ar and As as sides is det(A) times
the signed area of parallelogram with r and s as sides.
10.73 As described in the text, det(T) is the signed area of the parallelogram with Ti and Tj as sides. Since
multiplying by T is equivalent to a reflection, the area of the parallelogram formed by Ti and Tj has the same
magnitude as the area of the parallelogram formed by i and j, which is 1. But since multiplying by T is a reflection,
the orientation of the parallelogram is reversed, so the area of the parallelogram formed by Ti and Tj is —1.
Therefore, det(T) = -1.
10.74 Let A = ^ \ From A2 = I, we have det(A2) = det(I) = 1. But det(A2) = [det(A)]2, so det(A) = ±1,
which means ad - be = ±1.
We know that A is invertible, because A A = I. Hence, A-1 = A. But
ad - be \-c a J
a d -b
If ad - be = 1, then we must have - A 1 = " J, so a = d, b - -b, c - -c, and d - a. Hence, b = 0
and c - 0. Substituting into ad -be = 1, we get a2 - 0 = 1, so a = ±1.
'a b\ . _i f-d b
If ad-be = -1, then we must have = A 1 = ^ ^J, so a - -d, b = b,c- c, and d - -a. Substituting
Kc dj \ c -a
into ad - be - 1, we get -a2 - be - -1, so be - 1 - a2.
j2. If b = 0, then 1 - a2 = 0, so a = ±1 and c can be anything. If
b ± 0, then we can solve for c to get c = (1 - a2)/b. Hence, the solutions to A2 = I are of the form
where a and c are any real numbers, and b is any nonzero real number.
10.75 Expanding, we get (A - B)(A + B) = A(A + B) - B(A + B) = A2 + AB - BA - B2, which is equal to A2 - B2

if and only if AB - BA, i.e. the matrices A and B commute. Therefore, we do not have A2 - B2 = (A - B)(A + B)
for all matrices A and B. (See if you can construct a specific example in which A2 - B2 and (A - B)(A + B) are not
equal.)
10.76 Rearranging A + B = AB gives AB - A - B = 0. Factoring the first two terms gives A(B -1) - B = 0. Adding
I to both sides lets us factor more:
A(B - I) - B + I = I o A(B - I) - I(B - I) = I «■ (A - I)(B -1) = I.
The equation (A - I)(B - I) = I means that A - I and B - I are inverses. We showed in the text that inverses
commute (that is, we have M_1M = MM-1 for any invertible matrix M), so (B - I)(A - I) = I. This expands as
BA - B - A + I = I,
so BA = A + B = AB.
10.77 Since A Q j = Q, we have A2 Q j =aQ. But

-1
A2 = (A-51) 3 =A 3 -S
15 -10
205
fa + 3 b\ (A a + 5 b\
Now, let A = Then, we have A and A
\c + 3d) \A c + 5d)'
so the equations A and A give us
fa + 3 b\ (Aa + 5b\ _ ( -1 \
and
\c + 3d) \4c + 5d) ~ V-107 ‘
Solving the system of equations a + 3b = 4 and Aa + 5b = -1, we get a = -23/7 and b = 17/7, and from the
system of equations c + 3d = 5 and Ac + 5d = -10, we get c = -55/7 and d = 30/7. Therefore,
10.78
0 1 0 1W0 1 0 0
(a) Let A . Then A2 = 0. Hence, if A2 = 0, A is not necessarily equal to 0.
0 0 o oMo 0 0 0
(b) Let A A. First, we claim that A2 - (a + d)A + (ad - be)I = 0. Expanding, we get
a b
A2 - (a + d) A + (ad - be) I =
c dj \c“ d)-(a+d)(c ?
a2 + bc ab + bd a2 + ad ab + bd\ fad-be 0
+
ac + cd bc + d2) \ac + cd ad + d2J ' V 0 ad - be
a2 + be - a2 - ad + ad - be ab + bd - ab - bd
ac + cd-ac- cd be + d2 - ad - d2 + ad - be
0 0
0 0
= 0.
Given A3 = 0, we take the determinant of both sides to get det(A3) = det(0) = 0. But det(A3) = [det(A)]3,
so [det(A)]3 = 0. Therefore, det(A) = 0, so ad-bc - 0. Hence, the equation A2 - (a + d)A + (ad - be)I = 0
becomes
A2 - (a + d) A = 0,
which means A2 = (a + d)A. Multiplying both sides by A, we get A3 = (a + d) A2, so (a + d)A2 = 0.

If a + d ± 0, we can divide both sides by a + d, to get A2 = 0. If a + d — 0, then d = -a, and
2 _ fa2 + be ab + bd\ _ ( a2 + be b(a + (~a))\ _ fa2 + be 0 \ _ 2

- \ac + cd bc + d2) ~ \c(a + (-a)) bc + (-a)2) ~ V 0 a2 + bc) ~ ^ +
Let x = a2 + be, so A2 = xl. Multiplying both sides by A, we get A3 = xA, so xA = 0. If x t 0, we can divide
both sides by x, to get A = 0, and so A2 = 0. If x = 0, then A2 = xl = 0. In either case, A2 = 0.
206
CHAPTER 11 1
l Vectors and Matrices in Three Dimensions, Part 1
11.1.1
(a) If u - 3w = 2v, then w = *=&■ = (^+2j-k)-2(-i+5j+2k) = 6i-^-5k = 8: _ 5k

2i - |j -
(b) We seek constants a and b such that a(4i + 2j - k) + b(—i + 5j + 2k) = 0. Rearranging the left side gives
(4a - b)i + (2a + 5b)) + (-a + 2b)k = 0. Therefore, we must have 4a = b, 3a - -5b, and a - lb. Substituting
a = 2b into 4a = b gives 8b = b, from which we find b = 0 and a = 0. Therefore, there are not nonzero scalars
a and b such that au + bv = 0.
(c) We have
xu + yv = x(4i + 2j — k) + y(-i + 5j + 2k) = (4x - y)i + (2x + 5y)j + (—x + 2y)k.
We claim that this vector cannot be equal to k.

If (4x - y)i + (2x + 5y)j + (-x + 2y)k = k, then 4x - y = 0, 2x + 5y = 0, and -x + 2y = 1. The first equation
gives y = 4x, and substituting this into the second equation gives 2x + 20x = 0. This gives us x = 0, and
y = 4x then gives y = 0. But x = y = 0 does not satisfy -x + 2y = 1, so there do not exist scalars x and y such
that xu + yv = k.
1
11.1.2 If x is orthogonal to , then 0, which gives 3 - x + 4y = 0.
If x I is orthogonal to , then = 0, which gives -4 + x + 2y = 0.

X
Hence, we have x - 4y = 3 and x + 2y = 4. Subtracting the first equation from the second, we get 6y = 1, so
y = 1/6. Substituting, we get x - 4/6 = 3, so x = 11/3. The only solution is (x, y) = (11/3,1/6)
11.1.3 Let 6 be the angle between the two vectors. Then
-3 - 16 + 14 -5
cos 9 =
yj?>2 + (-4)2 + 22 ^/(-l)2 + 42 + 72 V29V66 Vl914'
207
CHAPTER 11. VECTORS AND MATRICES IN THREE DIMENSIONS, PART 1
so 0 = arccos( y=|) « [9T_
11.1.4 Since ||u|| is the magnitude of u and p is the distance from the point to the origin, we have ||u|| = p. From
here, we can finish either algebraically, or geometrically.
Geometric finish. By the definition of the ^-coordinate in spherical coordinates, the angle between u and k
is cft. We therefore have u • k = ||u|| ||k|| cos cp - ||u|| cos (p, so cos <p - jMjj, which means (p = arccos (3)
Algebraic finish. We have z-p cos cp = ||u|| cos cp, so cos <p = To get z in terms of u and k, we note that
u • k = (xi + yj + zk) • k = z. Hence, cos (p = tMt, so cp - arccos (3)
11.1.5
(a) Geometrically speaking, the equation ||v|| = 4 tells us that the length of v is 4, so (x, y,z) is 4 units from the
origin. Therefore, the graph is a sphere with radius 4 and center (0,0,0)
Algebraically speaking, we can note that ||v|| = ||xi + yj + zk|| = \Jx2 + y2 + z2, so we must have
yjx1 + y2 -I- z2 = 4. Squaring both sides gives x2 + y2 +z2 - 42, which is the equation of a sphere with radius
4 and center (0,0,0).
(b) Geometrically speaking, the equation ||v - w|| = 6 tells us that the vector from (1, -2,4) to (x, y, z) has length
6. In other words, (x, y, z) is 6 units from (1, -2,4), which means that it is on the
sphere with radius 6 and center (1, -2,4)
Algebraically speaking, we have
11 v - w|| = ||(xi + yj + zk) - (i - 2j + 4k)|| = ||(x - l)i + (y + 2)j + (z - 4)k|| = yj(x - l)2 + (y + 2)2 + (z - 4)2,
so we must have \J(x- l)2 + (y + 2)2 + (z - 4)2 = 6. This is the equation of a sphere with radius 6 and center
(1,-2,4).
(c) The origin is on the graph, since both sides of the given equation are 0 when v = 0. Otherwise, let 0
between the angle between v and w. We know that v • w = ||v||||w|| cos 0, so the given equation becomes
2||v||||w|| cos 0 = ||v||||w||. We know that ||w|| t 0, and assuming that ||v|| t 0 (i.e. v t 0), we can divide both
sides by ||v||||w|| to get 2cos 0 = 1, or cos 0 = \.
From cos 6 - \, we know that 0 is | or (or a multiple of 2n greater or less than these values). But
0 is the angle between v and w. Hence, the graph is a cone , where the apex is the origin, the axis has
direction vector w, and each line on the cone passing through the origin makes an angle of | with w.
11.2.1
( 4 • (-1) + (-1) • 4 + 3 • (-5) \
w (s i i) (i) (_1)' (~1) + 0 • 4 + 5 • (-5)

V(-3) • (-1) + (-1) ■ 4 + 6 • (-5)7
f 0-0 +(-3)-(-3)+ 2-3 \
?:) (f) (-1) • 0 + (-2) • (-3) + 7-3

\ (-3)-0 + 2-(-3)+ 1-3 7
=
208
Section 11.3
11.2.2
11.2.3 Let M = jd_. Then MA = (fcl)A = /c(IA) = k.A for any matrix A.
11.2.4 In the text, we found T by finding a matrix such that the product TA is formed by moving the first two
rows of A down one row, and moving the last row of A to the first row of the product. This isn't the only way we
can reorder the rows! We could also have moved the bottom two rows of A up one, and moved the top row to the
last row. Performing this operation three times gives us A back again. So, we seek the matrix U such that Ui = k,
Uj=i, and Uk = j. So, the columns of U, in order, are k, i, and j. Checking, we see that
011 012 013 \ / 021 022 023

021 022 023 = 031 032 033
031 032 033/ \flll 012 013
Notice that U = T2, where T is the matrix we found in the text such that T3 = I. So, we have
U3 = (T2)3 = T6 = (T3)2 = I2 = I,
and clearly U =£ I.
There are many other matrices that fit the problem; see if you can find others.
11.3.1
1 5 0
-2 6 -3 -3 -2
(a) -3 -2 6 = 1 -5 +0
7 -3 -1 -1 7
-1 7 -3
= l[(-2) • (-3) - 6 • 7] - 5[(—3) • (-3) - 6 • (-1)] + 0
= 1(—36) - 5(15)
-111
-3 4 0.5 0.25 -6
-6 -1 0.25 -1
0.25 -6 -1 = -3 -4 + 0.5
(b) 8 2 0 2 0 8
0 8 2
= -3[(-6) • 2 - (-1) • 8] - 4[0.25 • 2 - (-1) • 0] + 0.5[0.25 • 8 - (-6) • 0]
= -3(-4) - 4(0.5) + 0.5(2)
= 11 .
11.3.2 We have
an U\2 013 \ fb\\ b\2 b\3\ / an + bn 012 + ^12 «i3 + b\3

A+B = 021 022 023 + b2l b22 b23 J = I «21 + ^21 «22 + ^22 «23 + ^23
.031 032 033 / \hl b32 b33J \031+^31 «32 + &32 033 + ^33,
so
/011 + b\\ 021 + ^21 031 T b3\ \ /011 021 031 \ fb\\ b2\ b3i
(A + B)T = 012 + b\2 022 + b22 032 + b32 = 012 022 «32 J+( ^12 ^22 b32 = AT + Br.
, 013 + b\3 023 + ^23 033 + ^33/ \013 a23 033/ V ^13 ^23 ^33,
209
11.3.3 Since A and B have the same second and third columns, the matrices AT and BT have the same second
and third rows. Therefore, as showed in the text, we have det(Ar) + det(BT) = det(AT + BT). We also have
det(AT) = det(A) and det(BT) = det(B), so det(A) + det(B) = det(Ar + BT). In the previous Exercise, we showed
that At + Bt = (A + B)t, so we have
det(A) + det(B) = det(Ar + Br) = det((A + B)T) = det(A + B).
11.3.4 Subtracting twice the second row from the first row, we get
30 51 20 0 3 0
15 24 10 = 15 24 10
-46 28 -29 -46 28 -29
Expanding the determinant along the first row, we get
0 3 0
15 10
15 24 10
-46 -29
-46 28 -29
Adding three times the first row to the second row, we get
15 10 15 10
—3[15 • 1 10 • (-1)] = -3(25) =
-46 -29 -1 1
an ai2 013
022 023
11.3.5 Expanding the determinant along the first column, we get 0 «22 023 - 011 = 011022033
0 033
0 0 033
11.4.1
( (3)(-3) + (4)(7) (3)(4) + (4)(8) (3)(-2) + (4)(-l) (3)(-l) + (4)(0) \

(-3 4 -2
(—!)(—3) + (5 )(7) ( 1)(4) + (5)(8) (—!)(—2) + (5)(-l) (-1)(-1) + (5)(0)
V 7 8-1
V (0)(—3) + (2)(7) (0)(4) + (2)(8) (0)(—2) + (2)(—1) (0)(-l) + (2)(0) )
/19 44 -10 -3\

38 36 -3
1
\14 16 -2 0/
11.4.2 Q (-1 5 -3) (m--i)

Vd)(--i)
(4)(5)
(1)(5)
(4)(--3)\
(1)(-■3)/
7-4
V-l
20
5
-12\
-3)
11.4.3 We have v w = V\W\ + V2W2 -i-h v„wn = w\V\ + W2V2 H-h wnvn = w • v.
11.4.4
(a) We have
v • (aw) = v$azv$ + v2(axv2) + • • • + vn(awn) = a(v\W\ + v2w2 + • • • + v„wn) = av • w,

(av) • w = (avi)wi + (av2)102 + • • • + (avn)wn - a{p\W\ + V2W2 + H vnwn) = ax ■ w,
so v • (aw) = (av) • w = av • w.
210
Review Problems
(b) We have
U • (v + w) = Ui(V\ + W\) + U2(V2 + W2) +-h Un(vn + Wn)
= (U\V\ + U\IV\) + (U2V2 + U2W2) H-+ (UnVn + U„Wn)
- U\V\ + U2V2 +-h UnVn + U\W\ + U2W2 H-+ UnWn
- U • V + U • W.
11.4.5 In order for IA to be defined, I must have 2 columns, since A has 2 rows. Since the product IA must equal
A, the product must have 2 rows. Therefore, I must have 2 rows. This suggests that we try using the 2x2 identity
matrix as I. Sure enough, we find that
/I 0\ /flu «i2 #i3\ _ fa 11 #12 #13 \
\0 1/ \fl21 fl22 #23/ V#21 #22 #23/
However, is this the only possible I? Letting I = , we have
(V #12 #13\ _ f#11 #12 #bA

\r sj \fl2i #22 #23/ \#21 #22 #23/
From the entry in the first column and first row of the product, we have pflu + qa2\, which can only equal flu if
p = 1 and q - 0. Similarly, we can show that r - 0 and s = 1, so the 2x2 identity matrix is the only matrix I such
that IA = A.
Review Problems
11.29
(a) We have ||u|| \J22 + 32 + (-5)2 V38 y/42 + 52 + (-1)2 = V42
f 2•2+3•4 \ f 2-3-4 \ /16\ /-10\

(b) (2u + 3v) • (u - 3v) = ( 2-3 + 3- 5 ]•( 3-3-5 ) = ( 21 )•( -12 j
\2 • (-5) + 3 • (-1)/ V"5 - 3 • (-!)/ V-13/ V -2 /
= 16 • (-10) + 21 • (-12) + (-13) • (-2) = -386
11.30 Checking the dot product of all possible pairs, we see that only the second and fourth vectors are
orthogonal:
-3 ] = (-3) • (-2) + 4 • (-3) + 2-3 = 0.
AA (vA
11.31 Let v = [ %2 ] and w = M/2 , and let 6 be the angle between v and w. Then v • w = ||v||||w||cos0.
V*3 / Vys/
Squaring both sides, we get (v • w)2 = ||v||2||w||2 cos2 0, so
(xiyi + x2y2 + x3y3)2 = {x\ + A + *3X1/1 + y2 + vl) cos2 6.
But cos2 d < 1, so (xiyi + x2y2 + *3ys)2 < (*1 + x\ + xl){y\ + y\ + y23).
211
11.32
/4 5 -2 -1\ /4 • (-1) + 5 • 1 + (-2) • 4

(a) -3 4 1 = 1 • (-1) + (-3) -1 + 4-4 ) =
Vo 1 -2 4 ) Vo- (-1) + 1 • 1 + (-2) • 4
(b) There are many 3x3 matrices that have the same property as in part (a). For example.
(c) Consider the equation
The determinant of the 3x3 matrix is
5 -2
-3 4 1 4
-3 4 = 4
1 -2 0 -2 + (-2)
1 -2
= 4[(—3) • (-2) - 4 • 1] - 5[1 • (-2) - 4 • 0] - 2[1 • 1 - (-3) • 0]
= 4(2) - 5(-2) - 2(1)
= 16.
Because this determinant is nonzero, the system of equations has a unique solution. In other words, the
/-7N
vector | 1 I is the only vector that can be multiplied by the matrix in part (a) to get 12
11.33
1 0 O'
(a) Let A = ( 0 0 0 ) and v Then
0 0 0
'10 0
Av = | 0 0 0
0 0 0
Thus, Av = 0 does not imply A = 0 or v - 0.

f\ 0 0\ /O 0 0\
(b) Let A = I 0 0 0 ) and B = ( 0 0 0 J. Then
\0 0 0/ \0 0 1
10 0
AB = | 0 0 0
x0 0 0
Thus, AB = 0 does not imply A = 0 or B = 0.
4 -2 -1
11.34 (0 5 -1
-2 0 -1
212
Review Problems
11.35
/ Pll Pl2 Pl3'

(a) Suppose that such a matrix P = | P21 P22 P23 ) exists. Then
\P31 P32 P33,
1 0 0N
P I 0 0 0
0 0 0
But
1 0 0\ / Pll P12 Pl3 >11 0 0N
P I 0 0 0 I = I p2i P22 P23 P21 0 0
.0 0 0/ \p3l P32 P33 VP31 0 0y
These two matrices cannot be equal, so no such matrix P exists.
(b) Let the columns of P be p3, p2, and p3, in that order. So, the first column of AP is Ap3. In order for this to
equal the second column of A for every A, we must have p3 = 1 , since this will cause each entry in
w
APl to be the corresponding entry in the second column of A. Similarly, the product Ap2 must be the first
column of A, so p2 = 0 ), and Ap3 is the third column of A, so p3 = [ 0 ].
/0 1 0N
So, we let P = I 1 0 0 |. Then
\0 0 1
All 012 013 011 fli2 fll3 012 011 013
021 022 023 P= 021 0-22 #23 022 021 023

031 «32 033, 031 032 033 032 031 033,
Hence, for any matrix A, AP is the matrix that results when we swap the first two columns of A.
11.36 We have (A + B)v = 0 for all vectors v. Since this holds for all vectors v, we can take v = i, j, and k, to get
(A + B)i = 0, (A + B)j = 0, and (A + B)k = 0, respectively. Therefore, we have A + B = 0, so A = -B.
11.37
-2 -4 1
3 2 -5 2 -5 3
-5 3 2 = -2 -(-4) +
(a) 1 8 -3 8 -3 1
-3 1 8
= -2(3 • 8 - 2 • 1) + 4[(—5) • 8 - 2 • (-3)] + [(-5) • 1 - 3 • (-3)]
= -2(22) + 4(-34) + 4 = -176
1/3 1/5 6
1 -3 5 1 4 5 4 -3
4-3 5 +6
(b) “ 3 6 -10 ~ 5 -1 -10 -1 6
-1 6 -10
= i[(-3) • (-10) — 5 - 6] — ~ [4 • (-10) - 5 • (-1)] + 6[4 • 6 - (-3) • (-1)]

3 o
= - • 0 - - • (-35) + 6-21 133

3 5
213
11.38 By Cramer's Rule,
-3 -2 5 1 -3 5 1 -2 -3
4-5 3 2 4 3 2-5 4
-12 3 -5 -1 3 -5 2 -1
x ' v - z =
1-2 5 1-2 5 1 -2 5
2-5 3 2-5 3 2-5 3
-5 2 3 -5 2 3 -5 2 3
These determinants are
-3 -2 5
-5 3 4 3 4 -5
4-5 3 = -3 (-2) +5 = -3(-21) + 2(15) + 5(3) = 108,
2 3 -1 3 -1 2
-12 3
1-3 5
4 3 2 3 2 4
2 4 3 -(-3) +5 = 15 + 3(21) + 5(18) = 168,
-1 3 -5 3 -5 -1
-5 -1 3
1 -2 -3
-5 4 2 4 2 -5
2-5 4 (-2) + (-3) = -3 + 2(18) - 3(—21) = 96,
2 -1 -5 -1 -5 2
-5 2 -1
1- 2 5
-5 3 2 3 2 -5
2- 5 3 -(-2) +5 -21 + 2(21) + 5(-21) = -84.
2 3 -5 3 -5 2
-5 2 3
Hence, x = 108/(-84) = -9/7 , y = 168/(—84) = -2 , and z = 96/(-84) = -8/7
11.39 Subtracting the first row from the second row, we get
x-4 x-3 x-2 x -4 x-3 x-2 x-4 x—3 x-2

x-1 X x + 1 = (x - 1) - (x - 4) x - (x - 3) (x + l)-(x-2) = 3 3 3
x+2 x+3 x+4 x+2 x+3 x+4 x+2 x+3 x+4
Subtracting the first row from the third row, we get
x-4 x—3 x-2 x-4 x-3 x-2 x-4 x-3 x-2

3 3 3 — 3 3 3 = 3 3 3
x+2 x+3 x+4 (x + 2) - (x - 4) (x + 3) - (x - 3) (x + 4) — (x - 2) 6 6 6
x-4 x-3 x-2 x-4 x-3 x-2

Then 3 3 3 =2 3 3 3
6 6 6 3 3 3
11.40
(a) We have seen that the entry in row i and column j of the product AB is the dot product of the 2th row of A
and the ;th column of B. This is equal to the entry in row j and column i of (AB)T.
By the same token, the entry in row j and column i of the product B7 A7 is the dot product of the jtb row
of B7 and the zth column of A7. But the jth row of B7 is the yth column of B, and the 2th column of A7 is the
2th row of A. Hence, the entry in row j and column i of B7 A7 is the dot product of the /h column of B and
the 2th row of A. This equals the entry in row j and column i of (AB)7, as mentioned above. Since this holds
for all possible values of i and j, we have (AB)7 = B7A7.
214
Challenge Problems
(b) Suppose A is an r x s matrix and B is an s X c matrix, so that AB is defined. Then, we note that BTAT
exists because BT is a c x s matrix while Ar is an s xr matrix. Next we note that our solution in part (a)
does not ever mention the dimensions of the matrices; every step holds for this part, as well. So, we have
(AB)t = BtAt.
11.41
4 0 2 ' / (4)(7) + (0)(0) + (2)(—1) (4)(—1) + (0)(-2) + (2)(2) \

5 -1 -1 (5) (7) + (-1)(0) + (-1)(-1) (5)(—1) + ( 1)( 2) + (—1)(2)
( 3
-1
-2
5
7
2
(3)(7) + (-2)(0) + (7)(—1)
\ (-1)(7) + (5)(0) + (2)(-l)
(3)(—1) + (—2)( 2) + (7) (2)
(-!)(-!) + (5)(—2) + (2)(2) J
11.42 In order for AB to be defined, the number of columns in A must equal the number of rows in B, so B has
4 rows. The number of columns in AB equals the number of columns in B, so because C = AB has 6 columns, B
also has 6 columns. Therefore, B is a 4x6 matrix.
11.43
(a) Let A be an n x n matrix and B be an m x m matrix. If AB is defined, then n - m, so B is an n X n matrix, and

the product AB is also an nxn matrix.
(b) No . Suppose A is an r x n matrix and B is an n x r matrix. Then, AB is an r X r matrix. If r t- n, then A and

B are not square, but AB is.
Challenge Problems
11.44 From the given information, we have ||u|| = ||v|| = ||w|| = 1 and u-v = u- w = vw = 0. Then
r • u = (au + b\ + cw) • u = an ■ u + bu ■ v + cu • w = a||u||2 = a.
Similarly,
r • v = (flu + bv + cw) • v = au ■ v + b\ • v + cv • w = fr||v||2 = b,
and
t ■ w = (flu + b\ + cw) • w = au • w + bv ■ w + cw • w = c||w||2 = c.
11.45
(a) Let 6 be the angle between u and v. Then u v = ||u||||v|| cos 0. But
u • v = (i + j + zk) • (2i - j + 3k) = 1 • 2 +1 • (-1) + z • 3 = 3z + 1,
lull = Vl2 + l2 + z2 = Vz2 + 2, and ||v|| = ^Jl2 + (-1)2 + 32 = Vl4. Therefore, we have
3z + 1 = 14(z2 + 2) cos 6.
For 6 = f, we have cos 6 = cos f = 0, so 3z + 1 = 0, which means z = -1/3
215
(b) For 9 = we have cos 9 = cos | so 3z + 1 = \J 14(z2 + 2) • Multiplying both sides by 2, we get
6z + 2 = y/l4(z2 + 2). Squaring both sides, we get (6z + 2)2 = 14(z2 + 2), which expands as 36z2 + 24z + 4 =
14z2 + 28, so 22z2 + 24z - 24 = 0, or llz2 + 12z ~ 12 = 0. By the quadratic formula, we have
-12 ± y/l22 - 4 • 11 • (-12) _ -12 ± V672 _ -12 ± 4 V42 _ -6 ± 2 V42

Z_ 2-11 "" 22 ~~ 22 ” 11
The negative root is extraneous, because it makes 3z + 1 negative, but our initial equation is 3z + 1 =
+ 2) • 1. Therefore, the only value of z that satisfies the equation is (-6 + 2 V42)/ll
(c) Let c = cos 9, so 3z + 1 = (\J 14(z2 + 2))c. Our goal is to find the largest value of c for which there is a real
value of z that satisfies this equation. Squaring both sides, we get
9z2 + 6z + 1 = 14(z2 + 2)c2 = 14c2z2 + 28c2.
Rearranging this equation gives (14c2 - 9)z2 - 6z + (28c2 - 1) = 0, which is a quadratic equation in z. So, we
seek the largest value of c such that this quadratic has real roots. This quadratic equation has a real root in
z if and only if its discriminant is nonnegative. The discriminant is
(-6)2 - 4(14c2 - 9)(28c2 - 1) = 36 - (1568c4 - 1064c2 + 36) = -1568c4 + 1064c2 = 56c2(19 - 28c2).
This quantity is nonnegative if and only if c2 < |§. Therefore, we can find a value of z such that the angle
between u and v is 9 if and only if cos2 9 < ||. So, the desired maximum value of cos2 9 is || .
11.46 Subtracting the second row from the first row, we have
1 a a2 0 a-b a2 -b2
1 b
1 b b2 lb b2 = -(a - b) + (a2 - b2)
1 c
1 c c2
We can take out a factor of a - b, to get
1 b2 1 b 1 b2 1 b b2 b
-{a - b) + (a2 - b2) = —(a - b) (a - b){a + b) = (a-b) „2 + (a + b)
1 c2 1 c 1 c2 1 c c
= (a - b)[-(c2 - b2) + (a + b)(c - b)] = (a - b)[(b2 - c2) - (a + b)(b - c)]
- {a - b)[(b + c)(b - c) - (a + b)(b - c)] = (a - b)(b - c)[(b + c) - (a + b)]
(a - b)(b - c)(c - a)
We also could have tackled this problem with a little clever insight into polynomials. Clearly, the determinant,
if nonzero, has at most degree 3, since every term in the expansion has degree 3. Moreover, since the determinant
has two equal rows if a = b, the determinant is 0 if a = b. So, we conclude that a-b is a factor of the polynomial
that is the determinant. Similarly, b-c and c - a are also factors of the polynomial, so the desired polynomial is
k(a - b)(b - c)(c - a) for some constant k. Choosing a = 0, b = 1, c -2 gives a determinant of 2, so we must have
2 -k{a- b)(b - c)(c-a) - k(0 -1)(1 -2)(2- 0), which gives k = 1, and the desired polynomial is (a - b)(b - c)(c - a) ,
as before.
11.47 If Mt = -M, then det(Mr) = det(-M) = (-1)3 det(M) = - det(M). But det(MT) = det(M), so det(M) =
- det(M). Therefore, det(M) = 0.
11.48
(a) If Av = Av, then Av - Av = 0, so (A - AI)v = 0. If det(A - AI) ^ 0, then the equation (A - AI)v = 0 has a unique
solution, which is clearly v = 0. Therefore, if (A - AI)v = 0 for some nonzero vector v, then det(A — AI) = 0.
216
Challenge Problems
(b) We have
8-A 1 3
det(A - AI) = -4 1-A 7
8 1 -1 - A
1-A 7 -4 7 -4 1-A
= (-8 - A) +3
1 -1 - A 8 -1 - A 8 1
= "(A + 8)[(1 - A)(-l - A) - 7 • 1] - [(—4)(—1 - A) - 7 • 8] + 3[(-4)(l) - (1 - A)(8)]
= -(A + 8)(A2 - 8) - (4A - 52) + 3(8A - 12)
= (-A3 - 8A2 + 8A + 64) + (-4A + 52) + (24A - 36)
= -A3 - 8A2 + 28A + 80.
This polynomial factors as -(A - 4)(A + 2)(A + 10) = 0, so the values of A such that det(A - AI) = 0 are
A = ~4, -2, and -10
(c) Let v = ( y ]. Then for each eigenvalue A from part (a), we want to find a vector v such that Av = Av, or
-8x + y + 3z = Ax,
-4x + y + 7z = Ay,
8x + y - z = Az.
Equivalently,
(-8 - A)x + y + 3z = 0,
-4x + (1 - A)y + 7z = 0,
8x + y + (-1 - A)z = 0.
For A = 4, this system becomes

-12x + y + 3z = 0,
-4x - 3y + 7z = 0,
8x + y - 5z = 0.
Subtracting the third equation from the first equation to eliminate y, we get -20x + 8z = 0, so 20x = 8z, or
5x = 2z. Let x = 2 and z - 5. Then from the first equation, we have y = 12x - 3z = 9. These values satisfy
all three equations, so for A = 4, we can take v -
For A = -2, this system becomes

-6x + y + 3z = 0,
-4x + 3y + 7z = 0,
8x + y + z = 0.
Subtracting the third equation from the first equation to eliminate y, we get -14x + 2z = 0, so 14x = 2z, or
z = 7x. Let x = 1 and z = 7. Then from the first equation, we have y = 6x-3z = -15. These values satisfy
all three equations, so for A = -2, we can take v =
217
Finally, for A = -10, this system becomes
2x + y + 3z = 0,
-4x + lly + 7z = 0,
8x + y + 9z = 0.
Subtracting the first equation from the third equation to eliminate y, we get 6x + 6z = 0, so z = —x. Let x = 1
and z = —1. Then from the first equation, we have y = —2x — 3z = 1. These values satisfy all three equations.
so for A = -10, we can take v =
11.49 Computing the first few powers of A, we find
1 1 0\ /T 2 l\ l 3 3'
A= 0 1 1 ,A2 = 0 1 2 , A3 = 0 1 3 , A - A • A3 , A5 = A-A'
4 _
0 01/ Vo 0 1 VO 0 1
It appears that
A" =
for all positive integers n. We prove this using induction.
/1 k
The claim is true for n - 1, so assume that it is true for some positive integer k, i.e. Aa = ( 0 1 . Then
Vo 0
,fc+i A • A*
e-k+2k\
k+1
1 A
Thus, the claim is true for n = k + 1, so by induction, it holds for all positive integers n.
11.50 Since A has 3 columns and B has 2 rows, M must be a 3 X 2 matrix. So, we must have
m ii mi2
3
m2 i m2 2
5
m3i m32
Equating the entries in the first column of the product to the corresponding entries in B gives us the two equations
5mn + 3m2i - 2m3i - -6,

-mu + 5m2i + 4m3i = 8.
We can't produce any more equations involving mu, mî, and m3i. Since we only have two linear equations for
these three variables, we cannot find a unique solution (if there is a solution). But we are able to find solutions.
218
Challenge Problems
For example, adding 5 times the second equation to the first, we get 28m2i + 18m3i = 34. We can't eliminate any
more variables, so we can choose any value for ni2i and then find corresponding ran and ra31. For example, if we
let mu = -2, we have I8/W31 = 34 - 28m2i = 90, so 77J31 = 5. From either of our original equations, we have ran = 2.
Turning to the second column of the product and the corresponding entries in B, we have
5ran + 3ni22 — 2ra32 — —20,

-mu + 5ra22 + 4ra32 -- 24.
We can again choose any value for one of the variables, and then find values for the other two. For example, if
mi2 = ~3/ we will find ni22 = 1 and m.1,2 = 4. Checking, we find that we do indeed have
2 -3'
5 3-2 -6 -20
-2 11 =
-15 4 8 24
5 4
2 -3\
so M = 1-2 1 satisfies the problem, but it is not the only matrix that does so. For example, the matrix
5 4/
-9 8 \
M = | 7 —8 j also satisfies the problem.
-9 18/
11.51
(a) Let A = Q and B = Then
AB =
and
Thus, AB = -BA. But det(A) = -1 and det(B) = 1, so both A and B are nonsingular.
(b) If AB = -B A, then det(AB) = det(-BA). Since-BA is a 3x3 matrix, det(-BA) = (-l)3det(BA) = -det(BA),
so det(AB) = -det(BA). We also have det(BA) = det(B)det(A) = det(A)det(B) = det(AB), so det(AB) =
- det(BA) gives us det(AB) = - det(AB), which means that det(AB) = 0. Since det(AB) = det(A) det(B), we
must have det(A) = 0 or det(B) = 0. In other words, at least one of A and B must be singular.
11.52 Let
bn b\2 b\3
,2006
bi\ b22 b23
, ^31 b32 bz3
Since all the entries of A are even, all the entries of A2006 are even, i.e. b,-y is even for all i, j. Then
l-bn ~b\2 ~b\3

det(I - A2006) = -&21 1 - &22 —^23
-&31 -I732 1 - &33
1 - &22 -&23 -&21 —^23 -&21 1 — &22

= (1 - bn) (-&12) + (-&13)
-b32 1 - &33 —^31 1 ~ &33 -&31 -&32
The first term expands as
1 - Z?22 ~bl 3
(1 ~bn) = (1 - &ll)[(l _ &22)(1 ~ ^33) _ (-&23)(“^32)]-
-&32 1 - &33
219
Since all of the bij are even, the expressions 1 - b\\, 1 - £>22/ and 1 - £>33 are odd and (—b23)(-b32) is even. Therefore,
the expression (1 - bn)[(l - b22)(1 - b33) - (-b23)(-b32)] is odd.
Since b32 and b\3 are even, so are
~b2\ -b23 -b2\ 1 - b22

(~bn) and (-&13)
~b3i 1 - b33 ~b3\ ~b32
Therefore, det(I — A2006) is the sum of one odd and two even terms, which means det(I - A2006) is odd. In particular,
it cannot be equal to 0.
220
CHAPTER 12
l Vectors and Matrices in Three Dimensions, Part 2
1
12.1.1 The vector from (-4,5,0) to a point on the line must be in the same or opposite direction as the vector
from (-4,5,0) to (-5, -1,2). Therefore, if we take
-5N -r
v = and w = 5
0
then the line consists of all points (x, y, z) such that

x\
y = v + wt.
These values of v and w are not unique; we can take v to be any vector to a point on the line, and w to be any
-1 '
nonzero scalar multiple of [ -6
12.1.2 We can write
'MHMMX-
so a vector that is parallel to the graph is . (Any nonzero scalar multiple of this vector is parallel to the
graph as well.)
12.1.3 V is the graph of
D-CMvMT1-
In order for (10, -4,3) to be on the plane, we must have
10 = -2 + 2s - 3f,
-4 = -1 - s,
3 = 4 + s + 2t.
221
From the second equation, we have s = 3. Substituting this into the third equation gives 1 = -2. The values
(s, 1) = (3, -2) also satisfy the first equation, so letting (s, 1) = (3, -2) does give (x, y, z) = (10, -4,3) in our equation
for P. This means P passes through (10, -4,3).
To show that P does not pass through (-1,5,2), we must show the equation for P cannot give (x, y, z) = (-1,5,2)
for any values of s and 1. If such values of s and 1 exist, we must have
-1 = -2 + 2s - 31,
5 = -1 - s,
2 = 4 + s + 21.
The second equation gives us s = -6. Substituting this into the third equation gives 1 = 2. However, the values
(s, 1) = (-6,2) do not satisfy the first equation. So, there are no values of s and 1 for which (x, y, z) = (-1,5,2), which
means the P does not pass through this point.
12.1.4 Let A be (-1,2,4), B be (3, -1,2), and C be (5,1,6). Then, the plane through A, B, and C is the plane through
—4 -4
A generated by AB and AC. We have
AB = § - A = 4i - 3j - 2k,
AC = C-A = 6i-j+2k.
^ —4—4
If xi + 7j + 10k is in a plane generated by AB and AC, then we must have
xi + 7} + 10k = sAB + tAC = (4s + 6f)i + (-3s - t)j + (-2s + 2f)k.
This gives us the system
x = 4s + 61,
7 = -3s - 1,
10 = -2s + 21.
The last two equations give us (s, 1) = (-3,2). Substituting these into the first equation gives us x = |~0~|.
12.2.1 As described in the text, the vector ( -3 ] is normal to the graph of 4x - 3y + 12z = 5. (Note that the
12
negative of this vector is also normal to the plane.) The unit vector in the direction of this normal is
Multiplying this vector by -1 gives the other unit vector that is normal to the plane,
222
Section 12.3
12.2.2 Let v - v-ii + u2j + v3k and let (xo, yo, zo) be a point on the plane. Then, we have niXo + n2yo + ^3^0 = d since
(*o, }/0f zo) is on the plane, and we have n\V\ + n2u2 + ^3^3 = 0 because n • v = 0. Adding yi\Xq + H2yo + W3Z0 = d and
n\V\ + YI2V2 + n3v3 = 0 gives
ni(x 0 + vi) + n2{y 0 + v2) + n3(z0 + v3) = d,
so (xo + V\, y0 + x>2, Zo + U3) is on the plane. The vector between (xo + V\, yo + V2, Zo + v3) and (xo, yo, zo) is v, as desired.
12.2.3 Let v = 3i + 2j - 2k and w = 4i - j + 8k. Then the projection of v onto w is
v w 3 • 4 + 2 • (-1) + (-2) • 8
-w = (4i - j + 8k) = + A.; _ isu
l|w||2W 42 + (-l)2 + 82 271 ^ 27J 27*
12.2.4 The vector I -1

2v
is normal to the graph of 2x - y + 3x = 5, as is any scalar multiple of this vector,
flk'
-k ].
3k
Hence, 2k = 3, -k = b, and 3k = c. Then k = 3/2, so b = -k = -3/2 and c = 3k = 9/2
12.2.5 The distance from (3, b, -5) to the plane x + 2y - 2z - 8 = 0 is
13 + 2b - 2 • (-5) - 8| _ 12b + 5| _ |2b + 5|

^/l2 + 22 + (-2)2 V9 3
If this distance is 8, then 12b + 5| = 24, so 2b + 5 = ±24. If 2b + 5 = 24, then b = 19/2, and if 2b + 5 — -24, then
b = -29/2. So the solutions are b - 19/2 and -29/2
12.2.6 Let vi, v2, v3 and V4 be four vectors in three dimensions. We have two cases to consider.
Case 1: vi, v2, and v3 are linearly dependent. Then, there exist constants a\, a2, a3, not all 0, such that a^vi + a2v2 +
a3\3 = 0. Therefore, we have fliVi + a2v2 + a3V3 + 0v4 = 0, so there are constants a\, a2, a3, a4, not all 0, such that
a\\ 1 + a2v2 + fl3V3 + a4v4 = 0, which means that v4, v2, V3 and v4 are linearly dependent.
Case 2: v4, v2, and v3 are linearly independent. Let v4 = x4i + y4j + z4k As shown in the book, if v4, v2, and V3 are
linearly independent, then the graph of all (x, y, z) such that xi + yj + zk is a linear combination of v4, v2, and v3 is
all of IR3. Specifically, (x4, y4,z4) is in the graph, so v4 is a linear combination of v4, v2, and V3. Therefore, we have
v4 = fliVi + fl2v2 + a3v3 for some constants a\, a2, a3. This means we have a\\ 1 + fl2v2 + a3V3 + (-l)v4 = 0, so v4, v2,
V3 and v4 are linearly dependent.
For any set of vectors v4, v2,..., v„ with n > 4, we use the fact that any set of four vectors is linearly dependent.
Because v4, v2, V3 and v4 are linearly dependent, there exist constants fli,fl2,«3,04, not all zero, such that fliVi +
a2v2 + a3v3 + fl4v4 = 0. So, we have a\V\ + fl2v2 + ^3 + fl4v4 + OV5 +-1- 0v„ = 0 and not all of fli,fl2,fl3,a4 are 0, so
vi,v2,...,v„ are linearly dependent.
12.3.1
i j k 4 -1 3 -1 3 4
(a) axb = 3 4 -1 i- j + k = 7i + j + 25k =
7 0 -1 0 -1 7
-1 7 0
i j k
7 -1 7
(b) b x c = -1 7 0 k = 21i + 3j + 29k =
-1 -4 -1
-4 -1 3
223
i j k
-1 3 -4 3 -4 -1
c x a = -4 -1 3 i- j + -Hi + 5j - 13k
4 -1 3 -1 3 4
3 4-1
12.3.2 Let A = -i + 3j + 4k, B = 2i + j - 5k, and C = -7i - j. Then
AB = B - A = (2i + j - 5k) - (-i + 3j + 4k) = 3i - 2j - 9k,

AC = C - A = (-7i - j) - (-i + 3j + 4k) = -6i - 4j - 4k.
A vector that is orthogonal to both these vectors is
i j k
AB x AC = (3i - 2j - 9k) x (-6i - 4j - 4k) 3 -2 -9
-6 -4 -4
-2 -9 3 -9 3 -2
i— j + k = -28i + 66j - 24k.
-4 -4 -6 -4 -6 -4
Hence, the equation of the plane is of the form -28x + 66y - 24z = d for some constant d. Substituting the point
(-1,3,4), we get d = -28(-l) + 66(3) - 24(4) = 130. Therefore, the equation of the plane is -28x + 66y - 24z = 130,
or dividing by -2, we get 14x - 33y + 12z = -65
12.3.3 We have uxv = uxwif and only if u x v - u x w = 0, so u X (v - w) = 0. Thus, we want to take u, v,

and w such that u is parallel to v - w. So for example, we can take u and v to be any nonzero vectors such that
u + v^0, and let w = u + v. Then
uxw = ux(u + v) = uxu + uxv = uxv,
as desired. For example, let u = i, v = j, and w = i + j. Then, we have u x v = i x j = k and u x w = i x (i + j) =

ixi + ixj = 0 + k = k, souXv = uXw but v^w.
12.3.4 Let A = -i + 3j, B = -2i + 5j + 6k, and C = 6i - 4k. Then
AB = B - A = (—2i + 5j + 6k) - (-i + 3j) = -i + 2j + 6k,

AC = C - A = (6i - 4k) - (-i + 3j) = 7i - 3j - 4k.
The area of aABC is 1(AB)(AC) sin 0. We also have ||ABxAC|| = (AB)(AC) sin 0, so the desired area is j||ABxAC
We have
i j k
AB x AC = (-i + 2j + 6k) x (7i — 3j - 4k) = -1 2 6 lOi + 38j - Ilk.
7 -3 -4
3 Vl85
Therefore, the desired area is \ ||AB x AC|| = |||10i + 38j - llk|| = \ + 382 + (-11)2 = \ V1665 =
_2_
12.3.5 Let 0 be the angle between u and v. Then u v = ||u||||v|| cos 0 and ||u x v|| = ||u||||v|| sin 0, so
(u • v)2 + ||u x v||2 = (||u||||v|| cos 0)2 + (||u||||v|| sin 0)2 = ||u||2||v||2 cos2 0 + ||u||2||v||2 sin2 0
= ||u||2||v||2(cos2 0 + sin2 0) = ||u||2||v||2.
Therefore, ||u x v||2 = ||u||2||v||2 - (u • v)2.
224
Section 12.4
12.3.6
(a) Let a = (a2 j, b = f j, andc = ( c2 ]. Then

W \b3J \c3i
i j k
b2 b3 b\ b3 bi b2
b x c = bi b2 b3 1- k = (b2c3 - b3c2)i + (&3d - bxc3)] + (bxc2 - &2Ci)k,
C2 C3 ci c3 j + Ci c2
ci c2 c3
and
i J k
a x (b x c) = al d2 d3
b2c3 - b3c2 b3Ci - bxc3 bxc2 - b2ci
a2 a3 fli a3 Cl\ u2
l -
b3cx - bxc3 bxc2 - b2cx b2c3 - b3c2 b\c2 - b2cx j + b2c3 - b3c2 b3Ci - bxc3
= {a2b\c2 - a2b2cx - a3b3cx + a3bxc3)i
- (axbxc2 - axb2cx - a3b2c3 + a3b3c2))
+ (U\b3c\ — axbxc3 - a2b2c3 + a2b3c2)k.
/ a2bxc2 + a3bxc3 — a2b2cx — a3b3cx\
- ( aib2Ci + a3b2c3 — axbxc2 - a3b3c2 I .
\fli&3Ci + d2b3c2 - axbxc3 - a2b2c3 J
Also,
bl\
(a • c)b - (a • b)c = (axcx + a2c2 + a3c3) \ b2 - (axbx + a2b2 + a3b3)
\b3 J \<~3 /
/ axbxcx + u2bxc2 + ci3bxc3\ / uxbxcx + u2b2cx + u3b3cx

= I axb2cx + a2b2c2 + a3b2c3 j — j uxbxc2 + u2b2c2 + u3b3c2
\axb3cx + a2b3c2 + a3b3c3 J \axbxc3 + a2b2c3 + a3b3c3
/a2bxc2 + a3bxc3 — a2b2cx - a3b3cx\
= axb2cx + a3b2c3 - axbxc2 - a3b3c2 .
\ n_ /■> _1_ _ /7„ Vi_ _ n~ /
Hence, a x (b x c) = (a • c)b - (a • b)c.

(b) Applying the result we proved in part (a), we have both bx(cxa) = (b a)c-(b c)aandcx(axb) = (c-b)a-(c-a)b
in addition to a x (b X c) = (a • c)b - (a • b)c. Adding all three equations, we get
a x (b x c) + b x (c x a) + c x (a x b) = [(a • c)b - (a • b)c] + [(b ■ a)c - (b • c)a] + [(c • b)a - (c • a)b] = 0.
12.4.1 The vectors emanating from the vertex (-2,1,5) that correspond to the edges of the parallelepiped are
225
Therefore, the volume of the parallelepiped is the absolute value of
6 -2 -7
1 1 -6 = -42,
6 -4 -5
or 42 .
fli fli fl2

(*' &3
12.4.2 We saw that if a = ( fl2 ), t> C2 ], then a • (b x c) bi hi b3 . Hence,
«3, \C3, Cl C2 C3
Cl C2 C3 fli a2 fl3 fli a2 fl3

c • (a x b) = fli fl2 fl3 = - Cl C2 C3 — bi b2 b3 = a ■ (b x c).
bi &2 C3 bi bi C3 Cl c2 C3
so (a x b) • c = c • (a x b) = a • (b x c).
Geometrically speaking, the absolute value of each expression equals the volume of the parallelepiped with
edges a, b, and c. If you picture yourself standing at the origin and looking along the interior diagonal of
the parallelepiped from the origin, you'll see that in both expressions, the vectors appear in the same order,
counterclockwise or clockwise, around the diagonal. Therefore, the two given expressions equal the same signed
volume of the parallelepiped.
12.4.3 Since each two rows are orthogonal, the parallelepiped with the rows as edges is a rectangular prism. The
volume of a rectangular prism equals the product of its dimensions, so the volume of the parallelepiped with the
rows as edges is the product of the magnitudes of the rows. This parallelepiped has volume | det(M)|, so | det(M)|
equals the product of the magnitudes of the rows.
12.5.1 WehaveAA 1 = I,sodet(AA / = det(I) = 1. We also have det(AA x) = det(A)det(A 1),sodet(A) det(A /
1, which means det(A) and det(A_1) are reciprocals.
12.5.2 A matrix does not have an inverse if and only if its determinant is zero. We have
(1 5 5\
det -2 6 x = 6 + 5x2 + 20 - (—2x) - (-10) - 30x = 5x2 - 28x + 36.
\ x -2 1/
Factoring and setting this equal to 0 gives (5x - 18)(x - 2) = 0. Therefore, the values of x for which the matrix does
not have an inverse are f and 2
12.5.3 No Note that
The addends on the left are singular, but their sum is not.
flu fli2 fli3

12.5.4 Let A = 0 fl22 fl23
0 0 fl33
226
Review Problems
(a) We have det(A) - flufl22fl33, so det(A) =£ 0 if and only if An, a22, and a33 are nonzero. A matrix is invertible if
and only if its determinant is nonzero, so an upper triangular matrix is invertible if and only if the entries
on its main diagonal are nonzero.
(b) Let M be the inverse of upper triangular matrix, so we must have
/mn mu mi3\ /an a12 «i3\ /I 0 0\
( m21 ni22 ni23 I | 0 U22 a23 ) = I 0 1 0 I .

\w31 m32 m33J \ 0 0 a33J \0 0 1/
The entry in the second row of the first column of MA is ra21An ■ Since MA = I, this entry must equal 0. But
flu is nonzero since A is an invertible upper triangular matrix. Therefore, m21 = 0. Similarly, the entry in
the third row of the first column of MA must be 0. This entry equals m3iAn and An t 0, so m3i - 0. Finally,
the entry in the third row of the second column of MA equals 0. This entry equals m3\a\2 + m32A22. Since
m3\ = 0 and #22 is nonzero, we must have m32 = 0. Combining this with m21 = m3\ = 0, we see that M is an
upper triangular matrix.
Review Problems
12.35
(a) The equation p • v = 4 becomes x + 2y - 5z = 4, which is the equation of a plane
(b) The equation p • v = p • w becomes
or x + 2y - 5z = -3x + 4y - 2z, which simplifies as 4x - 2y - 3z = 0. This is the equation of a plane
We also could have written p • v = p • w as p • (v - w) = 0, so the graph consists of all (x, y, z) such that
A
y is orthogonal to the constant vector given by v - w. Therefore, the graph is a plane through the origin
with v - w as a normal vector.

(c) If p • v = 4 and p • v = p • w, then it follows that p • v = 4 and p • w = 4. Hence, the graph in this part is
the intersection of the graphs in parts (a) and (b). The graphs in parts (a) and (b) are different planes, so
1
the graph must be a line . (The two planes must intersect, because the vector 2 is normal to the first
V —5 /
plane and the vector -2 is normal to the second plane, and these vectors are not linearly dependent.)
V-3/
12.36 The equation of the line passing through (2,4,2) and (1,6, -1) is given by
Hence, if (a, 2, b) lies on this line, then a = 2 - t, 4 + 2f = 2, and b = 2 - 3f. From the equation 4 + 2f = 2, we have
t = -1. Then a = 2 - t = [3] and b = 2 - 3t = [IF.
227
3\ / «
12.37 The vector nj = ( —2 is normal to the plane 3x — 2y + 8z = 5, and the vector n2 = b ) is normal to
.8 J V-4,
the plane ax + by - 4z = 7. These planes are parallel if and only if one of the vectors ni and n2 is a scalar multiple
of the other. The z-coordinate of ni is 8 and the z-coordinate of n2 is -4, so we must have ni = -2n2.
Comparing the x- and y-coordinates, we get 3 = -2a and -2 = -2b, so a = -3/2 and b = [~lj.
12.38
(a) The equation of the line passing through (5,4,1) and (11,1, -8) is given by
The equation of the line passing through (5, -1,2) and (0,14,7) is given by
Hence, if (x, y, z) lies on both lines, then s and t must satisfy the equations 5 + 6t = 5 - 5s, 4 - 3t = -1 + 15s,
and 1 - 9t - 2 + 5s. Multiplying the second equation by 2, we get 8 - 6t = -2 + 30s. Adding this to the
equation 5 + 6t = 5 - 5s, we get 13 = 3 + 25s. Solving for s, we find s = 2/5. Substituting into the equation
1 - 9t = 2 + 5s, we get 1 - 9t = 4. Solving for t, we find t = -1/3. Then
5-5s
and | -1 + 15s
2 +5s
so the two lines intersect at (3,5,4)
6 f'5\
(b) The line £ is parallel to the vector ( -3 , and the line m is parallel to the vector 15 . Then a vector that
rV Vs/
is normal to the plane containing both £ and m is given by the cross product of these two vectors, which is
i j k
-3 -9 6 -3
6 -3 -9 k
15 5 -5 15
-5 15 5
120i + 15 j + 75k
Hence, the equation oftheplane is of the form 120x+15y+75z = d for some constantd. Substituting the point
(3,5,4), we get d = 120(3)+ 15(5)+ 75(4) = 735, so the equation of the plane is given by 120x + 15y+75z = 735,
or dividing by 15, we get 8x + y + 5z = 49
12.39 Solution 1: Let P = (x, y,z), Pi = (xi, yi,Zi), P2 = (x2, y2,z2), and P3 = (x3, y3,z3). Then
x - Xi X2 - Xi X3 - Xi
pip= 1 y-yi pip2 = ( yz-yi pip3 = [ ys - yi

Z — Zi Z2-Z1 z3 - Zi
228
Review Problems
so
x-x\ y-yi z-z\
xz — X\ y2- yi z2 - Zi
X3 - X1 1/3 - y! Z3 - Zi
is the signed area of the parallelepiped whose edges are PiP, P\P2, and P1P3. This volume is 0 if and only if P lies
in the same plane as Pi, P2, and P3.
Solution 2: The determinant is clearly of the form ax + by + cz + d, where a, b, c, and d are constants, so the graph
of the equation formed when we set this determinant equal to 0 is a plane. We can show that this is the desired
plane if we can show that the determinant is 0 when (x,y,z) equals each of (xi,yi,Zi), (x2/y2,z2), and (x3,1/3,23)
(and therefore all three points are on the graph of the equation). When (x,y,z) - (x\,y\,Z\), the first row of the
determinant consists of three Os, so the determinant is 0. If (x, y, z) equals (x2, y2, z2), then the first and second rows
are the same, so the determinant is 0. If (x, y, z) equals (x3, y3, Z3), then the first and third rows are the same, so,
again, the determinant is 0. Therefore, the graph of the given equation is the desired plane.
12.40
(a) We have
/ 4 a-b\
I —a +2b J
\2 a + 3 b J
If this is equal to p = [ -4 ], then 4a - b - 9, -a + 2b = -4, and 2a + 3b - 1. Solving the system 4a - b = 9,
-a + 2b = -4 gives a = 2, b = —1. Note that these values of a and b satisfy the third equation, 2a + 3b = 1.
Therefore, the only solution is (a, b) - (2,-1)
(b) We will show that it is impossible to have two different pairs (a, b) of scalars that satisfy t = av + few. We
let (ai, bi) and (a2, b2) be pairs of scalars such that aiv + b\w = a2\ + b2w. We will show that we must have
(fli,&i) = (a2/b2).
Rearranging aiv+i>iw = a2v+b2w gives (a\-a2)v—(bi—b2)vt = 0. Since v and w are linearly independent,
we must have a\ - a2 = b\ - b2 = 0, so a\ - a2 and b\ - b2. Therefore, the pair of scalars (a,b) such that
r = av + bw is unique.
12.41
(a) If u • w = v • w, then u • w - v • w = 0, so (u - v) • w = 0. This holds for all vectors w, so in particular, we can

take w = u - v to get (u - v) • (u - v) = 0. But (u - v) • (u - v) = ||u - v||2, so ||u - v||2 = 0. Hence, ||u - v|| = 0,
which means u - v, or u = v.
We could also have noted that u i = v i, u j = v • j, and u • k = v • k, so each component of u equals
the corresponding component of v. Therefore, we have u = v.
(b) If u x w = v x w, then u X w - v x w = 0, so (u - v) X w = 0. Let t = u - v = so t x w = 0 for all
vectors w.
Taking w = i, we get
i j k
h h h h
txi = h h h 0 1 0 0
k = f3j - *2k.
1 0 0
This is equal to 0, so t2 = t$ = 0 Similarly, t X j = -f3i + Lk = 0, so t\ = 0. Therefore, t = 0, which means

U = V.
229
12.42 The cross product is equal to
i j k
2 -1 3 -1 3 2
3 2-1 l- j +
4 2 -3 2 -3 4
-3 4 2
= [2 • 2 - (-1) ■ 4]i - [3 • 2 - (-1) • (—3)]j + [3 • 4 - 2 • (-3)]k - 8i - 3j + 18k
12.43 Let v = ( y ]. Then
i j k
0 0 1 0 1 0
ixv 1 0 0 k = -zj + yk.
y z x z j + x V
x y z
If this is equal to k, then y = 1 and z = 0, but x can be any real number, and we have v = xi + j. Thus, i x v = k
does not imply v = j.
12.44 Let a =
(4-1A and b =
(~x\
0 J. Then a x b is orthogonal to both these vectors, so it is orthogonal to the
\2 / W
plane generated by these vectors. This cross product is equal to
i j k
-1 2 4 2 4 -1
axb = 4-12 l- j + k = -4i - 18j - k.
0 4 -1 4 -1 0
-10 4
Hence, the equation of the plane is of the form -4a -18y-z = d, where d is some constant. Since this plane passes
through the origin, we have d = 0. Therefore, the equation of the plane is -4x - 18y - z = 0, or multiplying by -1,
we get 4x + 18y + z = 0
12.45 We can express the parametric equations as
TGMv*
As discussed in the text, the graph of this equation is a line that has direction vector ( 1 ). If we take t - 0, then
-5,
(3 - (-2)\ / 5
we see that the point (3,4,0) lies on the line. The vector joining (-2,3,5) to (3,4,0) is 4-3 = 1
V 0-5 J \-5/
This vector and the direction vector are in the desired plane. The cross product of these vectors is orthogonal
to both vectors, so it is normal to the plane. This cross product is
i j k
1 -5 -2 -5 -2 1
-2 1 -5 i- j +
1 -5 5 -5 5 1
5 1 -5
= [1 • (-5) - 1 • (—5)]i - [(-2) • (-5) - (-5) • 5]j - [(-2) • 1 - 1 ■ 5]k
,4
= -35j - 7k
230
Review Problems
so the equation of the plane is of the form -35y — 7z = d, for some constant d. Substituting the point (-2,3,5), we
get d = -35(3) - 7(5) = -140, so the equation of the plane is -35y - 7z = -140, or dividing both sides by -7, we
get 5y + z = 20
12.46 Let a = 2i, b = j, and c = 2k. Then
a X b = (2i) x j = 2(i x j) = 2k = c.
and
b x c = j x (2k) = 2(j x k) = 2i = a,
but
c x a = (2k) X (2i) = 4(k x i) = 4j,
which is not equal to b.
12.47
(a) Let w = I w2 . Then

\W
i j k
-1 2 4 2 4 -1
v xw = 4 -1 2 i- k = (—2w2 - w3)i + (2zoi - 4w3)j + (zv\ + Aw2)k
w2 zv3 W\ w3 j + W\ w2
W\ w2 w3
/-2zv2 - w3\
= I 2wi - 4w3 .
\ W\ + 4^2 /
'-2w2 - w3\
Also, 2w\ - 4iv3 , so we can take M
w\ + Aw2 J
(b) The function /(w) = u x w satisfies f(kw) - kf(w) and /(v + w) = /(v) + /(w) for any scalar k and any
vectors v and w. (We proved these fundamental properties of the cross product in the text.) Therefore, / is
a linear function that maps vectors to vectors, so there is a unique matrix N such that /(w) = Nw for all w.
Let's see if we can find the matrix in terms of the components of u!
. Then
i ) k
u2 u3 Ml u3 U\ u2
U Xw= u2 ~ i-
Ml M3
xv2 W3 W\ w3 )+ W\ w2
W\ w2 m
-u3w2 + U2W3
= (-u3w2 + u2w3)i + (u3wi - u\w3)) + (—u2w\ + u\w2)k U3W\ - U\W3
—U2W\ + U\W2
-U3W2 + U2W3\ —m3 m2 \

(0
Also, u3W\ — U\W3 , so we can take N = M3 0 -Mi
-U2W\ + U\W2 J \-U2 Ml 0 )
12.48 In Exercise 12.3.5, we showed that ||u X v||2 = ||u||2||v||2 - (u • v)2. So if u • v = 0 and u x v = 0, then
|iu||2||v|i2 = 0, so ||u|| = 0 or ||v|| = 0. Therefore, u = 0 or v = 0.
231
12.49 Let O be the origin, and let U, V, and W be points such that
/«A Ah\ _ /w A
U = U-2 , V = i?2 , and W = w2 . The first determinant is the
\U3J \v3J \W3J__
volume of the shown parallelepiped with edges OU, OV, and OW. Let
this be parallelepiped P. Let X be the point such that X - V +U, so X is a
vertex of P. The second determinant in the problem is the volume of the
parallelepiped with edges OX, OV, and OW. Let this be parallelepiped
Q, and let Y be the vertex of Q such that OXYV is a face of <3.
Both OVYX and OVXU are parallelograms with OV as a base, and with height to that base equal to the distance
between parallel lines tf? and tj]}. Therefore, these parallelograms have equal area. Parallelograms OVYX and
OVXU are faces of Q and P, respectively. They have equal area, and they are in the same plane. Point W is a
vertex on the face opposite the respective parallelogram in both (3 and P, so the altitudes to these equal area faces
in the two parallelepipeds are equal. Since the two parallelepipeds have faces of equal area, and equal altitudes
to these faces, the parallelepipeds have equal volume, which means the given determinants are equal.
12.50 In each case, we find the inverse of the matrix by first figuring out the effect when a matrix M is multiplied
by the given matrix.
(a) Let the matrix be A. We have
'mu mn m13\ / m2\ m22 m23

m21 m22 m23 mu mu m3
m3i m32 m33 \m31 m32 m33.
Multiplying M by A s waps the first two rows of M. Multiplying AM on the left again by A swaps the rows
back, so A2M = M. Therefore, A2 = I, so A-1 = A. Checking, we have
/0 1 0\ /0 1 0\ /I 0 0\
1 0 01 0 0 =10 1 0 .
\0 0 lj \° 0 V V° 0 1/
(b) Let the matrix be B. We have
/0 0 1\ /mu mu mi3\ /m3i m32 m33\

BM =1 0 0 m21 17122 m3 = mu mi m3
\0 1 0/ \m3i m32 m33J \m21 m22 m23)
Multiplying M by B "cycles" the rows of M, moving the third row to the first row, and moving the other
two rows down one row. So, multiplying by B twice more gives us M back again. Therefore, we have
B3M = M, which means B3 = I. This gives us B2B = I, so B_1 = B2. Testing, we find that
/0 0 1\ /0 0 1\ /0 1 0\
B2= l 0 01 0 0 = 0 0 1,
\0 1 0/ \0 1 0/ V1 0 0/
and
(0 1 0\ /O 0 1\ fl 0 0\
b2b= o 0 11 0 0 = 0 1 0 .
\1 0 0/ V° 1 0/ \° 0 V
We also can note that if M is the inverse of B, then BM = I. Setting the result of the product BM equal
to I, we can read off the entries of M and find that
B-1 = M =
232
Challenge Problems
(c) Let the matrix be C. We have
m ii mn Wi3\ ( -mi -m32 -m33

CM = m21 m3
m2 2 = ( mu mi2 m3
0 1 0 / V mi m2 m-i V m21 m2 m23
So, multiplying M by C "cycles" the rows just like B in part (b) and also multiplies the new first row by
-1. Therefore, multiplying M by C three times gives -M, which means C3M = -M. So, multiplying by
C3 three more times gives us M back again: C6M = C3(-M) = -(-M) = M, which means C6 = I. Since
C5C = I, we have C_1 = C5.
We also could have equated our result of CM to I to read off the entries of M, and we find that
C-1
12.51 We seek a matrix M such that MAB = I. Multiplying both sides on the right by B 1 gives MABB 1 = IB 1,
so MA = B_1. Multiplying both sides on the right by A-1 gives MAA-1 = B-1A-1, so M = B_1 A Note that
this is not necessarily the same as A_1B _1, because matrix multiplication is not commutative
12.52 Let A be the invertible matrix and B be tire non-invertible matrix. Since B does not have an inverse, we have
det(B) = 0. Therefore, we have det(P) = det(AB) = det(A) det(B) = 0, which means P does not have an inverse
12.53 We showed in a Review Problem in thei ««««''«= 1A-

previous chapter that (AB)T = B' Ar. Letting B == A , we have
(AA_1)t = (A-1)tAt, so lT = (A_1)tAt. Since IT - I, we have (A_1)rAr = I, which means the iinverse of Ar is
(A-y.
Challenge Problems
12.54
M /a\
(a) Let v = fo and f = \ b . The point (a, b, 0) lies on the graph of z = 0, and the vector v - f is
W V
normal to the graph of z = 0, so f is the projection of v onto the xy-plane. Furthermore,
so we can take
M =
(b) Let v = ( b . We seek a vector f = f such that (s, t, u) lies on the graph ofx + y + z = 0, so s +1 + u = 0.
233
fa-s\
Also, we want v - f = b-t to be normal to the graph of x + y + z = 0. In other words,
\c-uj
for some scalar k, so a-s = k,b—t = k, andc-u = k. Adding all these equations, we get (a+b+c)-(s+t+u) = 3k.
But s + t + u - 0, so a + b + c = 3k, which means k - (a + b + c)/3.
Then
cn a+b+c 2a - b - c
a
II
II
1
1
3 3
a+b+c -a + 2b - c
t = b-k = b-
3 3
a+b+c -a-b + 2c
u = c-k = c-
3 3
so
Furthermore,
so we can take
12.55 There are many possible answers. We'll show how to find one pair of vectors u and w that satisfy the
problem. First, we can take
i j k -2 -2
u = vXi = 3 -2 1 k = j + 2k,
0 0
1 0 0
which is orthogonal to v.
Now, we want w to be orthogonal to both u and v. We can simply take
i j k
1 2 0 1
w = U XV= 0 1 2 k = 5i + 6j - 3k,
-2 1 3 -2
3-2 1
which is orthogonal to both u and v.
These aren't the only possible answers. We can let u be the cross product of v and any vector that is not a
scalar multiple of v, and then let w = u x v. By this process of constructing u and w, we guarantee that they are
orthogonal to each other, and orthogonal to v.
234
Challenge Problems
12.56 Let
All -A21 A3I \ / 0ii 012 013
P= -A12 A22 -A32 | ( 021 022 #23
det(A)
A13 -A23 A33 / 032 033
We wish to show that P = I. First, we confirm that pn = p22 = p33 = 1. We have
Pn det(A) (flli^n ~~ fl21^21 + fl3i^3i)/

1
P22 det(A) (_fll2^12 + fl22^22 _ a3lA32),
1
p33 det(A) (fll3^13 ~~ fl23-^23 + 033A33).
Each of the expressions in parentheses above is an expansion by minors expression for det(A), where we are
expanding along the first, second, and third columns of A in the expressions for p\\, p22, and £>33, respectively.
Therefore, we have pn = P22 = P33 = 1.
We also have to show that each of the other entries in P is 0. We'll only show one here; showing that each of
the others is 0 can be done in the same way. We have
1 a22 a23 a\2 fli3 012 013

P12 - (012A11 ~ a22A2i + a32A3i) = 012 a22 + 032
det(A) det(A) 032 033 032 0-33 022 023
1
(012(022033 _ 023^32) — 022(012033 _ 032013) + 032(012023 ~ 022013))
det(A)
= 0.
Multiplying out each of the other non-diagonal entries produces similar convenient cancellation that shows that
each of these equals 0, so P = I.
12.57 Let the vector be v = xi + yj + zk. The angle between v and i is a, so v • i = ||v||||i|| cost* = ||v|| cos a. But
v • i = x, so x = ||v|| cos a. Similarly, we find that y = ||v|| cos and z = ||v|| cos y. We then have
||v||2 = x2 + y2 + z2 = ||v||2 cos2 a + ||v||2 cos2 jS + ||v||2 cos2 y - ||v||2(cos2 a + cos2 j8 + cos2 y).
Since ||v||2 ± 0, we can divide by ||v||2 to give cos2 a + cos2 jS + cos2 y = 1.
12.58 Let w = ||v||u + ||u||v. Let a be the angle between u and w, and let jS be the angle between v and w, where
0 < a,p < 7i. Then
u • w _ u • (llvllu + ||u||v) _ ||v||u • U + ||u||u • V _ 11 v| 11 |u| [2 + IH|u- V _ ! |u| 111 v| I + U • V

cos« — ||u||||w|| ~ ||u||||w|| l|u||||w|| l|u||||w|| ||w||
Similarly,
v w V • (llvllu + | |u| | v) _ llvllu • V + ||u||y • V _ llvllu • V + llulllMI2 u V + IMIHvll
COS^ _ IMIllwll ~~ ||v||||w|| “ ||v||||w|| l|v||||w|| ||w||
Thus, we have cos a = cos ft. Since 0 < a, < n, we must have a - j8, so w bisects the angle between u and v.
12.59
(a) Taking a = b = v in (ii), we get v x v = -v x v, so v x v = 0.

(b) If v = v\\ + v2j + v3k and w = W\i + zv2j + w3k, then repeatedly applying (iv) gives
v x w = (z?ii + v2j + v3k) x (w\\ + w2j + w3k)

= X i) + V\W2(i X j) + ViW3(i x k)
+ ^2î(j x i) + v2w2(j x j) + v2w3(j x k)
+ v3w\{k x i) + v3w2(k x j) + v3w3(k x k).
235
From part (a), we have ixi = j x j = kxk = 0. From (i) and (ii), we have j xi = —ix j = —k, kxj = -jxk = -i,
and i X k = -k x i = —j. Hence,
vxw= x i) + v\W2(i x j) + V\W3(i x k)

+ V2Wi(j x i) + V2W2Q x j) + V2W3Q x k)
+ V3W\{k X i) + t^W^k X j) + ^3W3(k x k)
= V\W2k - V1W3) - U2Wik + V2W3I + V3IV1) - V3W2I
- (V2W3 - V3W2)l - (V1W3 - V3W-1)) + (V\ZV2 - f2î)k

v2 V3 V\ V3 V\ v2
i- j+
IV 2 IV3 W\ zv3 W\ IV 2
(c) Using the same notation as in part (b), since v x w = (V2W3 - v3W2)i - (^W3 - v3W\)) + (p\W2 - V2’W\)k,
(v X w) • V = Vi(V2Ws - V3W2) - V2(v-[W3 ~ V3W1) + V^{V\W2 ~ V2W\)
- V\V2l03 - V1V3W2 - V1V2IV3 + U2^3î + V1V3W2 - V2V3ZV\
= 0.
We then have
(v x w) • w = (—w x v) • w = -(w X v) • w = 0,
where we use the result we proved at the start of this part in the final step. Hence, vxwis orthogonal to
both v and w.
12.60 From the identity a x (b x c) = (a • c)b - (a • b)c (proved as an Exercise in Section 12.3), we get
(a x b) x (a x c) = [(a x b) • c]a - [(a x b) • a]c.
But a x b is orthogonal to a, so (a x b) • a = 0. Furthermore, as proved in an Exercise in Section 12.4, we have

(a x b) • c = a • (b x c), so
(a x b) x (a x c) = [(a x b) • c]a - [(a x b) • a]c = [(a x b) • c]a - (0)c = [(a x b) • c]a = [a • (b x c)]a.
12.61 Let e = a x b, so (a x b) • (c x d) = e • (c x d). From the identity a • (b x c) = (a X b) ■ c,
e • (c x d) = (e x c) • d = [(a x b) x c] • d.
But (a x b) x c = -c x (a x b) = (a • c)b - (b ■ c)a, so
[(a x b) x c] • d = [(a • c)b - (b • c)a] • d = (a • c)(b • d) - (a • d)(b • c).
12.62 Solution 1: Geometry. The expressions t>, and are

the projections of v onto a, b, and c. Consider the parallelepiped with V
these three vectors as edges, as shown at right. Because each two of a,
b, and c are orthogonal, this parallelepiped is a right rectangular prism
(or box), and v is a space diagonal of the box. From face OADB of the
box, we have D = projav + projbv, and from rectangle OCVD, we have
v = D + projcv = projav + projbv + projcv, as desired.
Solution 2: Algebra. Since a, b, c are linearly independent, there is a

unique triplet of constants (a, (5, y) such that
v = aa + jSb + yc.
236
Challenge Problems
Since ba — c-a — 0(we are given that a is orthogonal to b and c), we have
v • a = (aa + /3b + yc) • a = aa ■ a + /3b • a + yc • a = a ||a|
Therefore, we have a = Similarly, we have /3 = ^ and y = so
v•a v • b.t v c
v= ra + rb + —«-c.
12.63 We prove that no such matrix M exists by contradiction. Let vi, V2, and V3 be linearly independent vectors
in three dimensions. If we can find a vector u for each vector v such that Mu = v, then there exist vectors ui, U2,
and U3 such that Mui = vi, MU2 = V2, and MU3 = v3. Since ui, U2, and U3 are in two dimensions, they are linearly
dependent. Therefore, there are some constants c\, c2, C3, not all 0, such that C1U1 + C2U2 + C3U3 = 0. Multiplying
each of our equations with M by the respective c„ and then adding all three equations, we have
CiMui + C2MU2 + C3MU3 = C1V1 + C2V2 + C3V3.
Therefore, we have
C1V1 + C2V2 + C3V3 = M(ciUi + c2u2 + C3U3) = MO = 0,
which contradicts the assumption that vi, V2, and V3 are linearly independent. Therefore, no such matrix M exists.
12.64 We know how to find the area of the the triangle with the three points as vertices, and how to find the
distance between each pair of the points. We can combine these to find any altitude of the triangle. Let v be the
vector pointing from (0,1,2) to (4,1, -3), and let w be the vector pointing from (0,1,2) to (-2,5,1), so v = 4i - 5k,
and w = —2i + 4j - k. The area of the triangle with the three points as vertices is \ ||v x w||. We have
j k
0 0
v xw= 4 0 -5 k = 20i + 14j + 16k.
4 4
-2 4 -1
Therefore, the area of the triangle is |||v x w|| = \ V202 + 142 + 162 = \ V&52 = V213.
But the area. A, of this triangle is also given by A = \bh, where b is the distance between the points (4,1, -3) and
(-2,5,1), and h is the distance between the point (0,1,2) and the line through (4,1, -3) and (-2,5,1). The distance
between the points (4,1, -3) and (-2,5,1) is b = \J[4 - (-2)]2 + (1 - 5)2 + (-3 - l)2 = y/68 = 2 Vl7. Hence,
2A _ 2V2l3 _ 2V213 _ /213

b b 2^^Y7 V 17
As an alternative solution, we also could have found a vector from a point on the line to (0,1,2) such that the
vector is normal to the line. The length of this vector is the desired distance. We let r be the vector from (4,1, -3)
to (0,1,2), and let d be the direction vector of the line. Then, the difference r - projdr is normal to the line, and
goes from the line to (0,1,2). The magnitude of this difference is the desired distance.
12.65 Let b = AB, c = AC, and d = AD. From the given information about the volume of the parallelepiped with
edges AB, AC, and AD, we have
/ b\ &2 bo,
det ci c2 c3 | = 80.
\d\ d2 do,
We then relate each of the desired volumes to this volume.
237
(a) Since AD is the interior diagonal of P, AD = d is the sum of the vectors corresponding to the three edges of
—^ ^
P emanating from A. Two of these vectors are AB = b and AC - c, so the third vector is d - b - c. Hence,
the volume of P is the absolute value of
bi b2 b3
Cl C2 C3
d\ - b\ - c\ d2-b2- c2 d3 — b3 — C3
Adding the first row to the third row, we get
bi b3 bi b2 b3
Cl C2 C3 — Cl C2 C3
di-bi- ci d2 — b2 — c2 d3-b3 — c3 dx - ci d2 - c2 d3 - c3
Adding the second row to the third row, we get
bi b2 b3 b2 b3
H
Cl C2 C3 = Cl C2 c3
di — Ci d2 — c2 d3 — c3 di d2 d3
We know that the absolute value of this determinant is 80, so the volume of P is 80
(b) Let u, v, and w be the vectors corresponding to the edges of Q emanating from A. If b is the face diagonal
of the face formed from v and w, then b = v + w. Similarly, we can set c = u + w and d = u + v. Adding all
three equations, we get b + c + d = 2(u + v + w). Then u + v + w = b+^+d, which makes
b+c+d , -b + c + d
u = (u + v + w) - (v + w)
■ b—2—
Similarly, v = b ^+d and w = bT; d. Therefore, the volume of Q is the absolute value of
—&i +ci +di —bz^Cz^dz ~b3+c3+d3

2 2 2
h —ci+rfi + —Cz+dz b3~C3+d3
2 2 2
(12.1)
bi+c-j-di b3 +C, ~d'2 bz+cz-dz
2 2 2
Each entry in the first column is a linear combination of b\, c\, and d\. Similarly, each entry in the second
column is a linear combination of b2, c2, and d2, and each entry in the third column is a linear combination
of b3, c3, and d3. So, we can express this matrix as the product of two matrices whose determinants we can
find easily:
-bi +ci +d\ —b2+C2+d2 —bi+ci+dz \ /I 1
2 2 2 \ ( 2 2 b2 b3\
b\ ~C\ -\-d\ bz-cz+dz b3-c3+d3 1 _ 1 1 1
2 2 2 2 2 C2 c3 J ,
b\ +Ci — d\ &2“b^2—^2 b2+C2~d2 J \ 1 1 ^2 d3 J
we have
-bi +c 1 +di —^2"b^2"b^2 —bz+cs+dz 1 1 1
2 2 2 2 2 2 b\ b2 b3 -1 1 1 bi b2 b3
b\ ~C\ +d\ b2~C2+^2 b3—c3+d3 1 1 1
2 2 2 2 2 2 Cl C2 c3 1 -1 1 Cl c2 c3
b\ +Ci ~d\ b2+C2~d2 ^2"b^2— ^2 1 1 1 d\ d2 d3 1 1 -1 d\ d2 d3
2 2 2 2 2 2
The determinant of the first matrix is 4, and the absolute value of the second is 80, so the volume of Q is
40 .
238
Challenge Problems
We also could have manipulated the determinant on line (12.1), like we manipulated the determinant
in part (a). We add the first row to the second, and then to the third, to find
-b-\ +Ci +di —b2+C?+d? -bi+ci+di -b i +d +rfi ~b2+C2+d2 —b^+Cs+ds -bi+ci+<?i -b3+C3+d2
2 2 -^2 +0
2 2 2 2 2 2 2
h-O +di b2-c2+d2 h-ci+d*
2 2 2 <*1 rfl d-2 do,
b i +Ci — di b2+c2—d7 b2 +c2-d7 ^2~^^2~^2 Cl C2 C3
2 2 2 2 2 2
1
Factoring 2 out of the first row, then adding the second and third rows to the first gives:
-bi+ci+d-i —b2+C2+d2 ~b3 +C3+dl
2 2 2 1 b\- c\- d\ &2 — C2 — ^2 bo, ~ C3 — do, h f»2 ^3
1
dx dz do, d\ di do,
~ 9 d\ d2 do,
Cl C2 C3 Cl C2 C3 Cl C2 C3
Finally, exchanging the second and third rows gives
bx b2 h b\ \>2 bo,
dx d,3 = -l(-l) Cl C2 C3
Cl C2 C3 d\ d2 do,
and again we have a volume of 40.
12.66
(a) As explained in the text, such a matrix M exists if the function /(w) = projvw is linear. We have
(aw) • v w•v
/(aw) = projv(aw)v = -v = a- v = a(projvw) = a/(w)
and
r/ . . , . (u + w) ■ v u•v+w■v
/(u + w) = proiv(u + w) =-=—v =-^-v
Ml2 llvll2
u•v w•V ,.
= —-yV + —~2 ■v = pro)vu + pro)vw = /(u + w).
Since / is linear, there exists a matrix M such that /(w) = Mw. Therefore, there exists a matrix M such that
projvw = Mw for all w.
(b) The matrix M is not invertible, so det(M) = 0 . To see why this is the case, note that if M had an inverse M_1,
then multiplying both sides of the equation Mw = 0 by M_1 would give us w = 0 as the only vector whose
projection onto v is 0. However, the projection of any vector orthogonal to v onto v is 0, so the equation
Mw = 0 does not have a unique solution. Therefore, M is not invertible, which means its determinant is 0.
(c) Let w = Then the projection of w onto v is
3(3x-2y+5z)
38
w•v 3x - 2y + 5z 3x -2 y + 5z 2(3x-2y+5z)
-v = 38
32 + (-2)2 + 52 38
5(3x-2y+5z)
38
Also,
__6_ 3(3x-2y+5z) \
38 38 y+l2 38 \
_£ 2(3x-2y+5z) 1
38 _38X + 38^ “ 38Z 38 I
_10 10. 25. 5(3x-2y+5z) /
38 \x 38 y+ 38" 38 /
239
so we can take
9 6 15
38 38 38
6 4 10
M = 38 38 38
15 10 25
38 38 38
As an additional challenge, see if you can use the expression for M above to generalize this problem. That
is, find a matrix P in terms of a, b, and c such that for any vector v, the product Pv is the projection of v onto
ai + b] + ck. Here's a hint: we also could have solved this part by projecting i, j, and k onto v.
(d) A scalar multiple of v is its own projection onto v. Since Mw is the projection of w onto v, it's already a
scalar multiple of v. So, the projection of Mw onto v is simply Mw. Just for fun, let's check:
_ 6_ 9 6 / 342 228 570 _6_

15 \
38 38 38 38 \ ( 382 382 382 38
4_ 6 4 10 228 152 380 4_
M2 = 1 - 1 = M.
38 38 38 38 1 " 1 382 382 382 38
_10 15 10 25 J 570 380 950 _10
38 38 38 38 7 V 382 382 382 38
12.67 Let £>i and B2 be the points on lines ki and k2, respectively, such that B\B2 is the shortest distance between
the two lines. B\B2 must be perpendicular to both lines. To see why, suppose B2B2 is not perpendicular to k\.
Then, let the plane through B2 perpendicular to k\ meet k\ at C, so aB\CB2 is a right triangle with right angle at C.
Therefore, we have B2C < B\B2, so B]B2 is not the shortest distance between the two lines.
Since BiB2 is perpendicular to both lines, it must be parallel to vi x v2. To relate A\A2 to BiB2, V\, and v2, we
note that
A\A2 — A\B\ + B\B2 + B2A2.
—9 —>
Since A\ and Bi are on k\, and A2 and B2 are on k2r we have A\Bi = civi and A2B2 = c2v2. So, we have
A\A2 • (vi x v2) = (A]Bi + BiB2 + B2A2) • (vi x v2)

= (civi + BiB2 + c2v2) • (vi x v2)
= CiVi • (Vi x v2) + BiB2 • (vi x v2) + c2v2 • (V! x v2)

= 0 + B\B2 ■ (vj x v2) + 0.
Since BiB2 is parallel to vi x v2, we have
|AiA2 • (vi X v2)| = ||BiB2 • (vi X v2)|| = ||BiB2|| ||vi x v2||.
(We introduce the absolute value on the left to account for the fact that Bi B2 can be in the same or opposite direction
as vi x v2.) Since ||BiB2|| = B\B2, we have
|AiA2 • (V! x v2)|

BiB2
llvi x v2||
240
CHAPTER
I -1
Vector Geometry
.
13 1.1
(a) Let A be the origin. Since the midpoint of AC is the same as the midpoint of BD, we have (A+C)/2 = (B+D)/2.
. ^ ^ ^ ^ ^ ^
Since A = 0, we have C = B + D. But the head of B + D is the fourth vertex of the parallelogram with
consecutive vertices at B, A (the origin), and D, so ABCD is a parallelogram.
(b) Let ABCD be the quadrilateral. Let E be the midpoint of AB, F be the midpoint of BC, G be the midpoint
of CD, and H be the midpoint of DA. We will show that EFGH is a parallelogram by showing that the
midpoints of diagonals EG and EH coincide. If M is the midpoint of EG, then
a1 E+G ^ + A+B+C+D
M = --- = —---— = -;-—.
If N is the midpoint of FH, then
, f+H + A+B+C+D
N= 2 = 2 = -4-'
Since M - N, points M and N are the same. Because the midpoints of EG and FH are the same, the diagonals
of EFGH bisect each other, which means that EFGH is a parallelogram.
13.1.2 Suppose ABCD is a kite with AB = BC and CD = DA. To show that BD ± AC, we can show that BD-CA = 0.
The equal side lengths give AB • AB = BC ■ BC and CD • CD = DA • DA. Letting B be the origin simplifies the first
equation to A • A = C • C. We can write CD • CD = DA • DA as
(D-C)-(D-C) = (D-A)- (D - A).
Expanding both sides gives

D-D-2C-D + C-C = D-D-2A-D + A-A.
The D • D terms cancel and A • A = C • C eliminates two more terms and leaves -2C • D = -2A • D, which rearranges
as D • (A - C) = 0.
Looking back at the equation we wish to prove, we see that when B is the origin, the equation BD ■ CA = 0 is
D • (A - C) = 0, which we have just proved above. So, from D • (A - C) = 0 (and the fact that B is the origin), we
have (D - B) ■ (A - C) = 0, which means BD ■ CA = 0. Therefore, the diagonals of kite ABCD are perpendicular.
241
CHAPTER 13. VECTOR GEOMETRY
13.1.3 We have
4MN = 4 (N-M) = 4 f j = 2B + 2D-2A-2C,
and
AB + AD + CB + CD = B-A + D-A + B-C + D-C = 2B + 2D-2A-2C,
so AB + AD + CB +CD = 4MN.
13.1.4 Let ABCD be the quadrilateral. The square of the length of the median from the midpoint of AB to the
midpoint of CD is
/A + B C+D\ fA+B C+D\
[—2 2~) re¬
setting this equal to the square of the length of the other median gives
/A + B C + D\ /A + B C+D\_/B+C A + dA /B + C A + D\
V 2 2 /V 2 2 )~{ 2 2~)\ 2 2 j'
We multiply both sides by 4 to get rid of the fractions:
(A + B - C — D) • (A + B - C - D) = (B + C- A~D)-(B + C- A- D).
We'll clearly have a lot of terms in common when we expand both sides. Rather than expanding right away, we
start with a little clever regrouping, rewriting the equation as
((B A)) ■ ((B - D) - (C - A)) = ((B -D) + (C - A)) ■ ((B - D) + (C - A)).

Now, the cancellation when we expand is clear. Expanding gives
(B-D)-(B-D)-2(B-D)-(C-A) + (C-A)-(C-A) = (B-D)-(B-D) + 2(B-D)-(C-A) + (C-A)-(C-A)-

■+ + A "4 -+ —4
Simplifying this equation gives 4(B - D) • (C - A) = 0, from which we deduce that DB • AC = 0, which means that
BD I. AC. All of these steps are reversible, so we have BD _L AC if and only if the medians of ABCD are equal in
length.
13.1.5 Letting C be the origin, we have CA2 + CB2 = A ■ A + B ■ B. We also have M = so
2CM* + ^=2/*ilWT±T + 2 (B “ A) • (B - A)
= ^(A-A + 2A-B + B-B) + -^(B ■ B - 2B • A + A ■ A)

-4-4 —4—4
= A-A + B-B.
Therefore, we have CA2 + CB2 = 2CM2 + 4^.
13.1.6 We are given that AB • CD = AC • BD = 0, so
(B-A)-(D-C) = (C-A)-(D-B).
Expanding both sides gives
—4—4 —4—4 —4 —4 —4—4 —4—4 —4 —4 —4 —4 —4—4
B-D-B-C-A-D + A- C = C- D- C- B- A- D + A- B.
Cancelling the common terms (and noting that B ■ C - C ■ B), we have
—4 —4 —4—4 —4—4 —4—4
B-D + A- C- C- D + A-B.
Rearranging this gives B-D-B-A-C-D + C- A = 0, from which we have (B - C) • (D - A) = 0, so CB • AD = 0.
—4—4 —4—4
Therefore, CB and AD are orthogonal, so BC and AD are orthogonal.
242
Section 13.2
13.2.1 Let the vertices be A, B, and C, and the centroid be G. We have
A + B + C_ 1
Therefore, the centroid of the triangle is (l,§,§)
13.2.2
(a) The Angle Bisector Theorem tells us that if D is on BC such that AD bisects ABAC, then BD/CD = AB/AC =
c/b. We showed in the text that if point D is on BC such that BD/CD = k, then D = So, we have
B+JC bB + cC
D =
1 + § b+c
(b) Continuing the reasoning from part (a), we see that if E and F are on AC and AB, respectively, such that BE
and CF are angle bisectors of A ABC, then E
i = l~^c and F = • Looking at the forms of D, E, and F, we
have a natural guess for E Suppose we let
aA + bB + cC
f =
a+b+c
Then, we have
^ r* bB + cC bB + cC - bA - cA
AD - D - A = —-A =
b+c b+c
uA + bB + cC aA + bB + cC - aA - bA - cA bB + cC - bA - cA
AI = I- A = -A-
a+b+c a+b+c a+b+c
Therefore, we have Al = -AD, which means I is on AD. Similarly, I is on each of the other angle bisectors
of aABC, and we have found the desired expression for E
13.2.3 We must show that

3(PA2 + PB2 + PC2) - AB2 - BC2 - CA2 > 0.
Applying vectors to the problem, we have
3(PA2 + PB2 + PC2) - AB2 - BC2 - CA2 = 3((P - A) ■ (P - A) + (P - B) ■ (P - B) + (P - C) ■ (P - C))
- (B - A) ■ (B - A) - (C - B) ■ (C - B) - (A-C)-(A-C)
= 3(P ■P-2P-A+A-A + P-P-2P-B + B- B + P-P-2P-C + C-C)
-B-B + 2A-B-A-A-C-C + 2B-C-B-B-A-A + 2C-A-C-C
= 9P-P-6P-(A + B + C)+A-A + B- B + C- C + 2(A ■ B + B ■ C + C ■ A)
= 9P • P — 6P • (A + B + C) + (A + B + C) • (A + B + C)
= (3P-(A + B + C)) • (3P-(A + B + C))
= ||3P-(A + B + C
243
This final expression is clearly nonnegative, so 3(PA2 + PB2 + PC2) - AB2 — BC2 — CA2 > 0, as desired.
We have equality when 3P — (A + B + C) = 0, which occurs when P = (A + B + C)/3. Therefore, equality holds
if and only if P is the centroid of A ABC
Note: We could have simplified the above calculations by letting P be the origin. We then find that
3(PA2 + PB2 + PC2) - AB2 - BC2 - CA2 = ||A + B + C||2,
and equality holds in the desired inequality if and only if A + B + C = 0. Therefore, (A + B + C)/3 = 0, so the origin
(which is point P) is the centroid, as before. (Note that we cannot deduce that the orthocenter is the origin from
A + B + C = 0. When H is the orthocenter of A ABC, we only have H=A+B+C- Oif the circumcenter is the
origin.)
13.2.4 We showed in the text that OG = OH/3, where G is the centroid of A ABC. Therefore, we have 30G = OJA,
so 3(G — O) = H — O, which means H = 3G - 20. Since G - (A + B + C)/3 for any A ABC, we have H = 3G - 20 =
A + B + C - 20, as desired.
13.2.5 Let the tetrahedron be ABCD such that we have
Adding these four equations gives
A+B+C+D- ^14^J .
Subtracting 3 times each of our first four equations from the equation above gives us the four vertices of the
tetrahedron:
A = (23,-25,14), B = (14,8,-22), C = (14,17,17), D = (-34,14,14).
13.2.6 We let the circumcenter of A ABC be the origin, O. Since U is the reflection of O over BC W
and OB = OC, the quadrilateral OBUC is a rhombus. Every rhombus is also a parallelogram, so
U - B + C. Similarly, we have V = A + C and W = A + B. From here we offer a couple solutions.
Solution 1: Guess! We have HA = A- U- A- B-C. Next, we check convenient points,

looking for one such that the vector from U to the convenient point is in the same (or opposite)
—4
direction as HA.
■4 —4 ~4
First, we try the circumcenter. We have HO = -B - C. No good; that's not a constant times
HA.
—4 -4 —4 -4 -4 ~4 —4
Next, we try the orthocenter of A ABC, point H. We have HH = A + B + C- B- C = A. Another loser.
Next, the center of the nine-point circle of A ABC, point T. We have HP = -B — C = Woohoo!
We found it! Similarly, T is on and W(^, so the lines are concurrent at the center of the nine-point circle of
A ABC.
Solution 2: Symmetry. If there is a point Z at which the three lines meet, the symmetry of the problem suggests
—4 -4 —4 ~4 —4 —4 —4 -4
that there is some constant k such that Z = k(A + B + C). As before, we have HA = A - B - C. We hope to find k
244
Challenge Problems
such that UZ is some multiple of UA. We have
UZ = Z-U = k(A + B + C)-{B + C) = kA + (k-l)B + (k- 1 )C.

If this is a multiple of A - B - C, then we must have ft: = -(ft: — 1), from which we have k = so point Z is the center
of the nine-point circle of aABC. Similarly, the center of the nine-point circle of A ABC is on all three lines.
Challenge Problems
13.19 We have
WX2 + YZ2 - XY2 - ZW2 = (X - W) • (X - VVO + (Z - Y) • (Z - Y) - (Y - X) • (Y - X) - (W - Z) • (W - Z)
= X-X-2W-X+W-W + Z-Z-2Y-Z + Y-Y

- Y ■ Y + 2X ■ Y - X • X - W ■ W + 2Z ■ W - Z ■ Z
= 2(X -Y-Y-Z + Z- W- W-X)
= 2(Y • (X - Z) - W • (X - Z))
= 2((Y - W) • (X - Z)) = 2WY • ZX.
13.20 Let ABCD be the quadrilateral. We have
AB2 + BC2 + CD2 + DA2-AC2 - BD2 = (B - A) ■ (B - A) + (C-B) • (C - B) + (D - C) • (D - C) + (A - D) • {A - D)

-(C-A)-(C-A)-(D-B)-(D-B)
-A-A + B- B + C- C + D- D + 2A ■ C + 2B ■ D
-2A-B-2B-C-2C-D-2D-A.
We'd like to show that this expression equals 0. Here, we might seem stuck, but thinking about what we'd like to
prove, that ABCD is a parallelogram, helps us get unstuck. ABCD is a parallelogram if and only if A + C = B + D.
We could prove that A + C- B- D = 0if we can show that || A + C - B - D|| =0. Now that we know what to look
for, we see look back at that huge expression above and see that
AB2 + BC2 + CD2 + DA2 - AC2 - BD2 = (A + C-B-D)-(A + C-B- D).
Therefore, we have AB2 + BC2 + CD2 + DA2 = AC2 + BD2 if and only if \\A + C - B - D|| = 0, which is true if and
only if ABCD is a parallelogram, and our proof is complete.
13.21 Let T be the midpoint of MN. We have M-(A + B)/2 and N = (C + D)/2, so
_ M+N _ A+B+C+D
~ 2 ” 4
We have P = (A + C)/2 and Q = (B + D)/2, so (P + Q)/2 = (A + B + C + D)/4 = T, which means T is the midpoint of
PQ. Similarly, we have R = (A + D)/2 and S = (B + C)/2, so (R + S)/2 = (A + B + C + D)/4 = f and T is the midpoint
of RS, as well. Therefore, fcPSl, P&, and £3 are concurrent at T.
13.22 Letting X be the origin, we have
2(AX2 + BX2 + CX2 + DX2) - (AB2 + BC2 + CD2 + DA2)
= 2 (A-A + B-B + C-C + D-D)-((B-A)-(B~A)
+ (C-B)-(C-B) + (D-C)-(D-C) + (A-D)-(A- D))
245
When we expand the dot products, all the terms of the form A ■ A cancel, and we are left with
-2(A - B + B- C + C-D + D-A) = -2(B ■ (A + C) + D ■ (A + C))

= -2(B + D) • (A + C).
Therefore, the original expression equals 0 if and only if one of the following conditions is met:
• B + D = 0. In this case, the origin is the midpoint of BD. The origin is the intersection of the diagonals, so
this means that AC bisects diagonal BD.
• A + C = 0. For the same reasons as in the previous case, this means that diagonal BD bisects diagonal AC.
• B + D and A + C are nonzero and orthogonal to each other. Since A and C are on the same line through
the origin, the vector A + C is parallel to this line. Similarly, the vector B + D is parallel to diagonal BD.
Therefore, if (A + C) ± (B + D), then AC _L BD.
13.23 We showed in the text that if H is the orthocenter of A ABC and the origin is the circumcenter of aABC, then
—^ ^ ^
we have H = A + B + C. But in this problem, the origin is not the circumcenter of A ABC, because A, B, and C are
not equidistant from the origin. Therefore, we cannot find the orthocenter in the way Mr. Assumption tried.
13.24 Let O be the origin, so A-A = B- B = C-C. We have M = (A + B)/2 and
A+C+M A + C+^p A B C 1
X =
= -^=2+6 + 3=6(M + B+2C)'
Therefore, we have
CM • OX = (M - C) • X = ^ • 7(A + B - 2C) • (3A + B + 2C)

2 6
= — (3A • A + B • B - 4C • C + 4A • B - 4A • C) .
Since A-A = B- B = C-C, we have 3A • A + B • B - 4C • C = 0, so
CM-OX=— (4A-B-1A-C) = -A-(B-C).
This quantity equals 0 if and only if A(3 _L The line through O that is perpendicular to is the perpendicular
bisector of BC, because O is the circumcenter of A ABC. Therefore, the condition that _L lx* means that A is on
the perpendicular bisector of BC, which means that AB = AC.
We still have to show that if AB = AC, then we have Svt _L 5^. If AB = AC, then AB • AB = AC • AC, which
means (B - A) • (B - A) = (C - A) • (C - A). Expanding and rearranging, and using the fact that B • B = C • C, we have
2A • B - 2A • C = 0, from which we have A • (B - C) = 0. (We also could have noted that AB = AC and OB = OC
mean that both A and O are on the perpendicular bisector of BC, so X(3 _L which means A • (B - C) = 0.) We
can then reverse our steps above to find that CM • X = 0.
Therefore, we have tfbX _L if and only if AB = AC.
13.25 We let the center of the circle be the origin. This point is also the circumcenter of each of the four triangles,
so the vectors to the vertices of the new quadrilateral are in question are
A + B + C, B + C + D, C + D + A, D + A + B.
246
Challenge Problems
Therefore the vectors along the four sides of the quadrilateral are
A + B + C-(B + C + D) = A-D = DA,

B + C + D-{C + D + A) = B- A = AB,
C + D + A-(D + A + B) = C-B = BC,
D + A + B - (A + B + C) = D - C = CD.
Each vector representing a side of the new quadrilateral is the same as a vector representing the corresponding
side of ABCD, so the each side of the new quadrilateral has the same length and is parallel to the corresponding
side of ABCD. Therefore, the quadrilaterals are congruent.
13.26
(a) Let O be the origin, so \\A\\ = ||B|| = ||C|| = R, which means
OH2 = (A + B + C) ■ (A + B + C) = A- A + B- B + C- C + 2(A - B + B- C + C-A) = 3R2 + 2(A ■ B + B ■ C + C ■ A).
We also have A ■ B - ||A|| ||b|| cos /AOB = R2 cos /AOB. The desired equation has cosines of the angles of
A ABC, so we must relate /AOB to the angles of A ABC.
If A ABC is acute, then /AOB is a central angle that subtends the same arc in which /ACB is inscribed.
So we have /AOB = 2/ACB, which means that cos /AOB = cos2C. If /.ACB is right, then AB is a diameter
of the circumcircle, and /AOB - 180°. This gives us cos /AOB = cos 180° = cos 2C once again. Finally,
if /ACB is obtuse, then /AOB subtends an arc of the circumcircle on which point C lies. In this case, the
measure of ZC is one-half the major arc AB, which has measure 360° - ACB = 360° - /AOB. So, we have
cos 2C = cos(360° - /AOB) = cos /AOB, since cos(360° - 6) = cos 6 for any angle 6.
In all of these cases, we have cos /AOB - cos 2C, so
A-B = ||A||||B|| cos /AOB = R2 cos2C.
Similarly, we have B ■ C = R2 cos2A and C • A - R2 cos 2B. Substituting into our expression for OH2, we
have
OH2 - 3R2 + 2(R2 cos 2C + R2 cos 2A + R2 cos 2B) = R2{3 + 2(cos 2A + cos 2B + cos 2C)).
(b) Applying the cosine double angle identity, we have
OH2 = R2(3 + 2(2 cos2 A - 1 + 2 cos2 B - 1 + 2 cos2 C - 1)).
Dividing by R2, we have 3 + 2(2 cos2 A+ 2 cos2 B + 2 cos2 C - 3) = Since ^ must be nonnegative, we
have
3 + 2(2 cos2 A + 2 cos2 B + 2 cos2 C - 3) > 0.
Isolating cos2 A + cos2 B + cos2 C gives the desired cos2 A + cos2 B + cos2 C > \.
13.27 Suppose there is a point V through which all three lines pass. Let M be the C
midpoint of AB. Since must be parallel to CX, the vector from M to V must be
in the same or opposite direction as the vector from C to X. Therefore, we must have
MV = k\CX for some constant fci. Since MV = V -M, CX = X-C, andM = (A + B)/2,
we have ^ ^
V - f -1 = Hi - A (i3.i)
247
Similarly, we must also have
? - | - \ = MX - A),
=h$-6),
for some constants fc2 and k3. From here, we present two solutions.
Solution 1: Solve for V. We can solve for V by adding all three equations to get
3V - A - B - C = (h + k2 + k3)X - k\C - k2A- k3B.

Solving for V gives
^ _ (k\ + k2 + k3)X + (1 — ki)C + (1 — k2)A + (1 - k3)B
~ 3
But what are k\, k2, and k3?
Substituting this expression for V back into Equation (13.1), we have

-> ->
(k\ + k2 + k3)X + (1 — k\)C + (1 — k2)A + (1 — k3)B A B
= k1(X-C). (13.2)
3 2 2
—^ ^ ^ ^
This equation will hold for all X, A, B, and C if the coefficient of each vector on the left matches the corresponding
coefficient on the right. This is the case if
k\ + k2 "t k3
= *i.
3
1 -fc2 1
= 0,
2
i -h i
= 0,
3 2
1 —■ fc2
= ~h-
3
The second and third equations give us k2 = k3 = The fourth equation then gives us k\ = These values
also satisfy the first equation. So, we have found constants k\, k2, and k3 such that Equation (13.2) holds for all X,
A, B, and C. Putting these values for the constants back into our original equations for V gives us
P-4.-1
2 2
—^
-> B C
*-2-2
2 2
We see that all three equations are satisfied if V = (-X + A + B + C)/2. Therefore, there is a point V such that all
three equations are satisfied, which tells us that the three lines described in the problem are concurrent.
248
Challenge Problems
Solution 2: Educated Guessing. We must prove that there is some vector V and C
some constants k\, k2, and k3 such that we have
B -> ->
2 =k\(X- C),
^=k2(X-A),
~=k3(X-B).
From the diagram at right, it looks like VM = |CX, and that MV is in the opposite direction of CX. So, it looks
like we 11 have VM = —|CX. It looks like V has a similar relationship to each of the other midpoints we care about
in the problem, so we guess that we have k\ = k2 - k3 = — We substitute these values into the equations and
then proceed as above to show that there is a V that satisfies all three equations for all X, A, B, and C.
13.28 We have D
B+C+D A+B+C
UV=V-U-
3 3
10-V
D+E+F A+E+F
YX = X - Y =
3 3
llD-X
Since UV — YX, we have UV || XY and UV = XY. Similarly, we

have VW = ZY and WX = UZ. Therefore, each two opposite sides of
UVWXYZ are parallel and equal in length.
13.29 We let the foot of the altitude from D to face ABC be the origin,
both because this point is a special point of A ABC (the orthocenter), and
because this gives usA-D = B- D- C- D = 0. From the given iBDC - 90°, we have BD ■ CD - 0, which gives
us (D - B) ■ (D - C) = 0. Expanding the product on the left, and using the fact that B ■ D = C ■ D = 0, we have
D ■ D + B • C = 0, so D ■ D - -B ■ C.
We can show that lADB = 90° by showing that AD ■ BD = 0. We have
AD ■ BD = (D-A)-0~B)-D-D-D-B-A-D + A-B.
Since D ■ B = A ■ D = 0, we have AD ■ BD = D ■ D + A • B. Substituting D ■ D = -B ■ C from above, we have

AD ■ BD = -B ■ C + A ■ B = B ■ (A - C). Since the origin is the orthocenter, B is parallel to the altitude from B to AC.
Therefore, B is orthogonal to AC, which means B ■ (A - C) = 0. This gives us AD ■ BD = 0, so /.ADB = 90°. We can
go through essentially the same steps to show that ZADC = 90° as well.
13.30 If it is possible for Harrison to construct such a hexagon, then let ABCDEF be the hexagon. Then, the
vectors to the six points given are in some order. The sum of the first, third, and fifth
of these equals the sum of the second, fourth, and sixth. Another way of saying this is that the centroid of the
triangle with vertices at the first, third, and fifth midpoints is the same as the centroid of the triangle with vertices
at the second, fourth, and sixth midpoints.
Looking at the six given points, we see that the x-coordinate of one of the points is odd, while the x-coordinates
of the other five points are even. So, no matter how we split the six points into two groups of three, the sum of
the x-coordinates will be odd and even in the other. Therefore, it is impossible to divide the six points into two
groups of three points such that sum of the x-coordinates in one group is the same as the sum of the x-coordinates
in the other.
249
13.31 We have
AX • CX - CB ■ AX = (X - A) ■ (X - C) - (B - C) • (X - A)
= X- X- X- C- A- X + A- C- B- X + B- A + C- X- C- A
= X-X-B-X-A-X + B-A
= (X - A) • (X - B) = AX ■ BX.
We have AX • BX = 0 if and only if AX _!_ BX or if X is A or B. Therefore, X is on the circle with diameter AB.
Conversely, if X is on the circle with diameter AB, then AX • BX = 0, and we can reverse the steps above to show
that AX CX = CB- AX.
13.32 Let the center of the circle be the origin. For all i with 1 < i <n, we have
PiX ■ Pfx = (X - Pi) • (X - Pi) = X • X + Pi- Pi - 2X • Pi.

^ 11 112
Since the P, are on the circle centered at the origin with radius r, we have P, • P, = ||P,|| = r2. Since the polygon is
regular, the sum of the P, is 0 (for the same reason that the sum of the nth roots of unity is 0). Therefore, we have
(PiX)2 + (P2X)2 + (P3X)2 + • • • + (P„X)2 = nX ■ X + {Pi ■ Pi + P2 • Pi + ■ ■ + ••• + ?„• P„) - 2X • (Pi + P2 H-b Pn)
= nX ■ X + nr2 - 2X • 0.
Since X • X is the square of the distance between X and the origin, and the origin is the center of the circle, C, we
have the desired
(PiX)2 + (P2X)2 + (P3X)2 + • • • + (P„X)2 = nir2 + CX2).
11 -i 11 —£ I
13.33 Let the circumcenter be the origin. Therefore, we must have ||A|| = ||H — A . Since H - A + B + C and
||A|| = ||B|| = ||C|| = R, where R is the circumradius, we have R = \\(A + B + C) - A\\ - \\B + C||. Squaring both
sides gives
R2 = ||B + C||2 = (B + C) • (B + C).
Expanding the dot product, and noting that B ■ B = C ■ C = R , we have
P2 = B • B + 2B • C + C • C = 2K2 + 2||B|| ||C|| cos LBOC = 2R2 + 2R2 cos LBOC.
We therefore have cos LBOC — Since 0 < LBOC < n, we have LBOC = We have LB AC = jLBOC because
LBAC is inscribed in the same arc that central angle LBOC subtends. Therefore, the only possible value of ZA is
n
3 •
13.34 Let the given points be A, B, and C, respectively. Any segment connecting two vertices of an octahedron
is either an edge or one of the three space diagonals connecting opposite vertices. We note that AB - BC - 6 and
AC = 6 V2, so AB and BC are edges and AC is a space diagonal. We can find the fourth vertex D of the octahedron
in the plane of A ABC by noting that ABCD is a square, so CD = BA. We have BA = (2 - 6)i + (6 - 4)j + (-8 - (-4))k =
-4i+2j-4k. SinceCD - BA = -4i+2j-4kandpointCis(4,8,0), we find that D is (-4+4,2+8, -4+0) = (0,10,-4)
Let E and P be the other vertices of the octahedron, and let M be the center of square ABCD. To find E and P,
we note that ME and ME are normal to plane ABCD, and in opposite directions. Therefore, the vector AM x BM
is in the same direction as one of ME and MF, and in the opposite direction of the other.
250
Challenge Problems
^ The center of square ABCD is the midpoint of AC, which is (3,7, -4). We then have AM = i + j + 4k and
BM = —3i + 3j. We then find
i j k
AM x BM = 1 1 4 —12i - 12j + 6k.
-3 3 0
Therefore, one of ME and MF is in the direction of —12i - 12j + 6k and the other is in the opposite direction.
Because the octahedron is regular, each of these vectors has length equal to half the length of a space diagonal of
the octahedron. Therefore, we have ||ME|| = ||MF|| = AC/2 = 3 V2. The vector —12i - 12j + 6k has direction
-12i - 12j + 6k = —12i - 12j + 6k _ -12i - 12j + 6k _ _2. _ 2. ^ 1^

|| —12i — 12j +6k|| (-12)2 + (-12)2 + 62 18 31 3J + 3
So, we can take
ME = 3 V2 (-|i - |j + ik) = -2 V2i - 2 V2j + V2k,
ME = -3 V2 (~|i - |) + |k^) = 2 V2i + 2 V2j - V2k,
Since M is (3,7, -4), we find that E is (3-2 V2,7-2 V2, -4 + V2) and F is (3 + 2 V2,7 + 2 V2, -4 - V2)
13.35 Let a = and b = b\ + &2*- The desired area equals ^|fl||b|sin0, where 6 is the angle formed by
the segments connecting a and b to the origin. We have seen expressions like this when working with the cross
product. So, let's view this problem in terms of vectors. Let A and B be the points («i, a2,0) and (b\, b2,0) in space,
and let O be the origin. The area of the parallelogram with OA and OB as sides is ||OA x OB ||. This is the same
parallelogram as described in the problem, so its area equals the desired area. We have
OA xOB|| = ||(fli&2 - fl2î)k|| = \a\b2 - a2bi\.
We also have
^(ab-ab) ((fli + a2i)(bi - b2i) - (fli - a2i){b 1 + b2i)) = - \-lci\b2i + la2b\i\ = \a\b2 - a2bi\,
which equals the expression we found earlier for the area of the parallelogram. Therefore, the area of the
parallelogram equals \\(ab - ab)|, as desired.
13.36 The perpendicularity suggests vectors. We let A be the origin. Since AX _L D Y, we have AX • D Y = 0, which
becomes X • DY = 0 when we let A be the origin. We have Y = (A + B + C)/3 = (B + C)/3 since Y is the centroid of
AABC and A is the origin. We also have X-(B + C + D)/3 because X is the centroid of A BCD. Therefore, X • DY = 0
becomes
We multiply both sides by 9 to get rid of the fractions and have
(B + C + D)-(B + C-3D) = 0.
Expanding the dot product on the left gives
B-B + 2B-C-2B-D + C-C-2C-D-3D-D = 0. (13.3)
251
From AB = 6, we have B ■ B = AB2 = 36. Similarly, we have C ■ C - AC2 - 82 = 64. Letting the desired AD be t, we
have D • D = t2, and Equation (13.3) becomes
100 - 3t2 + 2B ■ C - 2B ■ D - 2C ■ D = 0. (13.4)
We have BC2 = ||B - C||2 = (B - C) • (B - C) = B ■ B - IB • C + C ■ C = 100 - 2B ■ C. Since BC2 = AD2 = t2, we have
t2 = 100 - 2B • C, so 2B • C = 100 - t2. Similarly from BD2 = ||B - D||2 = (B - D) • (B - D), we find 2B ■ D = t2 - 28,
and from CD2 = ||C-D||2 = (C-D)-(C-D), wefind2C-D = f2 + 28. Substituting these into Equation (13.4) gives
100 - 3f2 + (100 - t2) - (f2 - 28) - (t2 + 28) = 0,
from which we find 6t2 = 200, so t = V200/6 - VlOO/3 = 10 V3/3 . Notice that the answer equals yj(AB2 + AC2)/3.
Is this a coincidence?
13.37 We let O be the origin, and we let r be the radius of the sphere. Because P and Q are opposite vertices of a
parallelepiped with edges PU, PV, and PW, we have
Q = P + (U - P) + (V - P) + (W - P).
Therefore, we have
||Q||2 = (P + (li - P) + (V - P) + (W - P)) • (P + (U - P) + (V - P) + (W - P))
= P-P + 2P-(ll-P) + 2P-(y-P) + 2P-(W-P)
(U-P)-(U-P) + (V-P)-(V-P) + (W-P)-(W~ P)

+ 2(U - P) ■ (V - P) + 2(V - P) • (W - P) + 2(W - P) • (U - P).
Each dot product on the last line is 0 because any two edges that share a vertex of the parallelepiped are
orthogonal. Expanding the remaining terms, we see that all terms of the form 2P • U appear twice, once added
and once subtracted, so they all cancel. Since \\U.\\ - ||V|| - j|w|| = R, we are left with
IIqII2H|u|N||k||2 + ||w!|2-2||p||2 = 3^-|!pT-

Therefore, all possible Q are a distance of 3R2 - ||P||2 from the origin. (As an extra challenge, show that every
point that is 3R2 - ||P|| from the origin can possibly be point Q.)
13.38 We use the intersection of the diagonals as the origin, and we use complex numbers. As usual, each
lowercase letter is the complex number corresponding to the point denoted by the respective uppercase letter.
y+z+o w+x+o 1,
§2~g\ =-^-3-= -(y + z-w-x).
We wish to show that the quotient (h2 - hi)/(g2 ~ gi) is imaginary. Unfortunately, h2 and hi are a lot harder to deal
with. We know that t)H2 _L ^ , so h2/(z - w) is imaginary. Can we find what this quotient equals? For that, we
turn to trigonometry.
252
Challenge Problems
We focus on zWOZ because ZXOY of AXOY has the same measure. X

So, if we can express h2 in terms of this angle, then the same angle
will appear in our corresponding expression for h\. We let the feet of
the altitudes from O and W be A and B, as shown. From quadrilateral
ZAH2B, we have /AH2B + /AZB = n. We also have /AH2B + lBH20 = n,
so ZBH20 = /AZB, which means AOH2B ~ aWZB (since ZWBZ =
LOBH2 = n/2). From this similarity, we have ^ , so
OB
OH2 = WZ- -~ = WZ cot ZWOZ.
WB
We also have t)H2 X WZ, so to get h2r we rotate w - z by |, and scale the
result by cot ZWOZ. Therefore, with the diagram oriented as shown, we have h2 = i(w-z) cot ZWOZ. Similarly, we
have h\ = i(y - x) cot ZXOY. We have ZWOZ = /.XOY, so we have h2-h\ = i(w + x — y — z) cot ZWOZ. Combining
this with our earlier expression for g2- g\, we have
h2-h\ i{w + x - y - z) cot ZWOZ i(w + x - y - z) cot ZWOZ

- —-t----=-z-= — 3i cot ZWOZ.
g2~gi y + z-zv-x) -^(w + x-y-z)
Since this quotient is imaginary, we have fci G2 X h2H2.
13.39 We let the centroid of ABCD be the origin. We therefore have A + B + C + D = 0. We wish to show that
£ + F + G + H = 0.
In order to use the information relating the altitudes of ABCD to EFGH, we note that the altitude from A to face
BCD is parallel to AE. Since DC x DB is normal to face BCD, this cross product is in the same or opposite direction
as AE. We choose the orientation of ABCD such that these vectors have the same direction. (If the orientation is
the opposite, the rest of the solution is essentially the same.)
We have
DC x DB = (C-D)x(B-D) = CxB-CxD-DxB + DxD = CxB-CxD-DxB + 0.
We also have ||DCxDB|| = 2[BCD]. Letting V be the volume of ABCD, we have ha[BCD]/3 = V,soha = 3V/[BCD].
Since ||AE|| = k/ha, we have
II Crll k - k^BCD^ - k\\DCxDB\\

" ha ~ 3V 6V
Since AE and DC x DB are in the same direction, we have
AE = —(C x B - C x D - D x B).
6V
Similarly, we can show that
BF= 4-(ACxAD) = -p-(C x D - A x D - C x A),

6V 6V
CG = —{BA x BD) = -^7 (A xD-AxB-BxD),
6V 6V
DH= -yr-,{CA x CB) = -^- (A x B - A x C - C x B).
6V bv
Taking the sum of these four equations gives us a lot of cancellation, and leaves
AE + BF + CG + DH= zL(-D xB-CxA-BxD-AxC).

bv
253
Since BxD = —D x B and AxC = -D X A, the four terms on the right above sum to 0 as well, and we have
AE + BF + CG + DH = 0.
Writing each vector as a difference of vectors gives
E-A + F- B + G- C + H- D = 0.
Since A + B + C + D = 0, we have E + F + G + H = 0, as desired. Therefore, the centroid of EFGH is the same as the
centroid of ABCD.
13.40 The interior angles of a hexagon sum to An, so if all the angles of E s4 D
hexagon ABCDEF are equal, then each is equal to ^
Let si = AB, S2 = BC, S3 = CD, S4 = DE, S5 = EF, and S& = FA.

Let a be the complex number corresponding to A, and similarly for the
other points. Without loss of generality, assume that hexagon ABCDEF
is oriented so that AB is parallel to the real axis, i.e. b - a is real.
Then we can let b — a = Si, for some real number Si. Since LABC = ^f,
the argument of c - b, which corresponds to side BC, is | greater than the argument of b - a, which corresponds
to AB. Therefore, letting £ = em/3, we have c - b = CS2 for some real number S2. Similarly, we have d - c = C2S3,
e - d — C3S4, e — f = C4S5, and / - a = C5S6 for some real numbers S3, S4, S5, and S6- Adding all of the equations
corresponding to the sides of the hexagon, we find
Si + 0>2 + C2Ss + C3S4 + C% + C5S6 — 0.

Since C = era/3, we have C3 = ent = -1/ which makes the above equation
si + Cs2 + C2s3 - s4 - Cs5 - C2s6 = 0,
which we can rearrange as

(si - s4) + C(S2 - s5) + C2(s3 ~ s6) = 0.
We could now substitute em/3 = (1 + i V3)/2 and do some algebra to finish, but that's not nearly as illuminating
(or as fun!) as reasoning geometrically.
Let p = si - S4, q = S2 - S5, and r = S3 - s^, so that our equation above is

p + Cq + - 0. Let P correspond to p in the complex plane, and suppose
Si — S4 > 0, so P is on the positive real axis, as shown. (The proof is essentially the
same if si - S4 < 0.) Let Q correspond to p + Op Writing the equation p + Cq + C2r = 0
as p + 0? = ~C2r, we see that Q also corresponds to -C,2r. Since r is real, the point
corresponding to -C,2r is on the line through the origin and the second and fourth
quadrants that makes and angle of tt/3 with the positive real axis. Likewise, since
q is real, the point corresponding to p + Qq must be on the line shown through P
that makes an angle of tt/3 with the positive real axis. (There are two lines through
P that make an angle of n/3 with the positive real axis; we take the one that does
not pass through the second quadrant, as shown.)
Point Q must be on both of these lines, so it is the shown intersection point. Letting O be the origin, we
therefore have ZPOQ = ZOPQ = tt/3, so ZOQP = n/3 and the triangle is equilateral. This gives us OP = PQ = QO.
Since OP = \p\, PQ = \Cq\ = \q\, and OQ = |-rC2| = |r||-£2| = \r\, we have \p\ = \q\ = \r\, which gives us \si - s4| =
|S2 - S5I = IS3 - Sg|, so |AB - DE| = |BC - EP| = |CD - FA\, as desired.
See if you can find another solution by proving that the equation p + C,q + C2^ = 0 implies that the triangle with
vertices p, Qq, and £2r is equilateral.
254
Challenge Problems
13.41 Let a - BC, c = AB, and b = AC. From Exercise 13.2.2, we have D = , £ = aJ^f, and F = . We
simplify the problem a bitby letting A be the origin. We choose A rather than one of the other vertices because
this choice gives ED and ED symmetric forms. We therefore have
bB + cC cC
ED = D-E = ((a + c)(bB + cC) - {b + c)cC)
b+c a+c (a + c)(b + c)
1
(abB + acC + bcB + c2C - bcC — c2C)
(a + c)(b + c)
1
(iabB + acC + bc(B - C)).
(a + c)(b + c)
Similarly, we have
FD = --—--(czLB + acC + bc(C - B)).

(a + b)(b + c)v v ”
We are given that ED ■ FD = 0, so
(flbB + acC + bc(B - C)) • (abB + acC + bc(C - B)) = 0.
Since the first two terms in each of the vectors in the dot product above are the same, but the last terms of each
are opposites, we have
(abB + acC + bc(B - C)) ■ (abB + acC + bc(C — B)) = (abB + acC) • (abB + acC) - b2c2(C - B) • (C - B).
Expanding this expression, and noting that B • B = AB2 = c2, C • C = AC2 = b2, and (C - B) ■ (C - B) = BC2 = a2, we
have
(abB + acC) ■ (abB + acC) - b2c2(C - B) -(C-B) = a2b2B • B + 2a2bcB ■ C + a2c2C ■ C - b2c2a2
= a2b2c2 + 2a2bcB ■ C + a2b2c2 - b2c2a2 - 2a2bcB ■ C + a2b2c2.
Setting this equal to 0 gives B ■ C = -\bc. Since B • C = be cos iBAC, we have cos iBAC = from which we see
that the only possible value of ABAC is 120°
To see that such a triangle is possible, consider an isosceles triangle ABC A

with ZBAC = 120°, so /.ABC = /ACB = 30°. There are many ways to show that
ZEDF = 90° in such a triangle. For example, by symmetry, we have EE || BC.
Then, since BE bisects ZB, we have ZFEB = /EBC = /FBE, which means ABEF
is isosceles with EF = BF. Similarly, ACEF is isosceles with EF = EC. Letting X D X
be the foot of the altitude from E to CD, triangle EXC is a 30-60-90 triangle. Therefore, EX = EC/2 = EE/2, which
means altitude DZ from D to EE also has length EE/2. Since EZ = FZ = EE/2 = DZ and DZ _L EF, triangle DZE is
a 45-45-90 triangle, so ZEDE = 2/EDZ — 90°.
13.42 Let co = enlD, a primitive 14th root of unity, so cou = 1. We place the C
diagram in the complex plane so that A, B, C, and D correspond to the complex
numbers a = 0, b = k, c = kco (so that AB = AC and /BAC = f), and d = 1, where
we take k to be a positive real number such that \b - c\ = 1.
Since /BAC = y and AC = AB, we have /ABC - /ACB = yL Since /ABC =

y1, we see that the angle BC makes with the x-axis on the right is 7i - y- = yL
Therefore, the argument of c - & is yL Since the magnitude of c - b is 1, we have c - b = co4. We also have
c-b-kco-k = k(co - 1), so k(co - 1) = o>4, which means
or
co - 1
255
ar a>5-m+1 _ |ai5-o)+l| As usual, we get rid of the magnitudes by

Then CD = \c - d\ = \kco - 1| = d)-l
1 co—1 - ku-l| '
considering the square of the desirec length:
rp2 _ l^5 ~ co + 1|2 _ (w5 - co + l)(a>5 - co + 1)

|<u - 1|2 (<n - 1)((U - 1)
We'll handle the numerator and denominator separately. But first, we note that co7 = em = -1 and co14 = e2m = 1,
and that cok - (co)k = (1 /co)k = l/cok = co14/cok = cou~k. Using these relationships, we find
(co5 — co + l)(a>5 — co + 1) = (co5 — co + l)(a>5 - co + 1) = (co5 — + l)(a>9 - co13 + 1) = (o>5 — + l)(-<u2 + co6 + 1)
= CO^(—CO2 + + 1) — co (—co2 + + 1) + 1 (—co2 + coP + 1)
= -co7 + co11 + co5 + co3 - co7 - co - co2 + co6 +1
— 1— co4 + co5 + co3+ 1- CO -co2 + co6+ 1
= co6 + co5 - co4 + co3 - co2 - CO + 3.
Similarly, we have
(co - 1 )(co — 1) = (co — 1 )(co - 1) = (co - 1 )(o>13 - 1)

= co14 - co13 - <u + l = l + o;6-a; + l = aj6-a; + 2.
At first, it doesn't look like we can do anything with the ratio of our expressions for \co5 - co + 1|2 and \co - 1|2. In
the past, we have simplified expressions involving roots of unity by deriving equations that are satisfied by those
roots of unity. We try the same here, starting with our earlier observation that co7 = — 1. Rearranging this gives
co7 + 1 = 0, and factoring gives
co7 + 1 = (co + 1 )(co6 - co5 + co4 - co3 + co2 - co + 1)
Since co + —1, we have co6 - co5 + co4 - co3 + co2 - co +1 = 0, which means co6 = co5 - co4 + co3 - co2 + co -1. Substituting
this into our expressions for \co5 - co + 1|2 and \co - 1|2 gives
(co5 - co + 1 )(co5 - co + 1) = co6 + co5 - co4 + co3 - co2 - co + 3
= (co5 - co4 + co3 - co2 + CO - l) + (O5 - co4 + co3 - co2 - CO +3

= 2co5 - 2co4 + 2co3 - 2co2 + 2 = 2(co5 - co4 + co3 - co2 + 1).
and
(co - 1 )(co -l) = cob-co + 2 = (co5 - co4 + co3 - co2 + co - 1) - co + 2
- co5
= , ,5 - co4 + co3 - co2 + 1.
Therefore, we have CD2 = —=---5-■=—— = 2, which means CD - V2

OlP — co* + co6 — coL + 1 -
256
'
1
www.artofproblemsolving.com
The Art of Problem Solving (AoPS) is:
• Books
For over 16 years, the classic Art of Problem Solving books have been used by students as a resource for
the American Mathematics Competitions and other national and local math events.
Every school should have this in their math library.

- Paul Zeitz, past coach of the U.S. International Mathematical Olympiad team
The Art of Problem Solving Introduction and Intermediate texts, together with our new Precalculus and
Calculus texts, form a complete curriculum for outstanding math students in grades 6-12.
The new book [Introduction to Counting & Probability] is great. 1 have started to use it in my classes on a
regular basis. I can see the improvement in my kids over just a short period.
- Jeff Boyd, 4-time MATHCOUNTS National Competition winning coach
• Classes
The Art of Problem Solving offers online classes on topics such as number theory, counting, geometry,
algebra, and more at beginning, intermediate, and Olympiad levels.
All the children were very engaged. It's the best use of technology I have ever seen.
- Mary Fay-Zenk, coach of National Champion California MATHCOUNTS teams
• Forum
As of November 2009, the Art of Problem Solving Forum has over 71,000 members who have posted
over 1,600,000 messages on our discussion board. Members can also participate in any of our free "Math
Jams."
I'd just like to thank the coordinators of this site for taking the time to set it up... I think this is a great site,
and I bet just about anyone else here would say the same...
- AoPS Community Member
• Resources
We have links to summer programs, book resources, problem sources, national and local competitions,
scholarship listings, a math wiki, and a ETgX tutorial.
I'd like to commend you on your wonderful site. It's informative, welcoming, and supportive of the math
community. I wish it had been around when I was growing up.
- AoPS Community Member
• ... and more!
Membership is FREE! Come join the Art of Problem Solving community today!
—'-■-
the Art of Problem Solving

Precalculus
Precalculus is part of the acclaimed Art of Problem Solving curriculum designed to challenge
high-performing middle and high school students. Precalculus covers trigonometry, complex
numbers, vectors, and matrices. It includes nearly 1000 problems, ranging from routine exercises
to extremely challenging problems drawn from major mathematics competitions such as the
American Invitational Mathematics Exam and the USA Mathematical Olympiad. Almost half
of the problems have full, detailed solutions in the text, and the rest have full solutions in the
accompanying Solutions Manual.
“Every time one of your books arrives on my doorstep I couldn Y be more impressed with what
you're doing... For the right kind of student—and I was one myself, I just wish these books had
been around when I was a kid—your books must be like a prayer answered. And as I’ve always
said to every math teacher I meet, this is the way to teach mathematics to everyone.”
Steve Olson, author of Count Down
Naoki Sato is a curriculum developer and the director of Art of Problem Solving’s Worldwide Online Olympiad Training
(WOOT). He won first place in the 1993 Canadian Mathematical Olympiad and is a 2-time medalist at the International
Mathematical Olympiad. He is a former Deputy Leader of the Canadian IMO team.
Richard Rusczyk is the founder of www.aitofproblemsolving.com. He is co-author of the Art of Problem Solving, Volumes
1 and 2, and Intermediate Algebra and author of Introduction to Algebra and Introduction to Geometry. He was a national
MATHCOUNTS participant, a three-time participant in the Math Olympiad Summer Program, a perfect scorer on the AIME,
and a USA Math Olympiad Winner.
ISBN 978-1-934124-17-8
00 >

AOPS Precalculus Solutions

Uploaded by

Copyright:

Available Formats

You might also like

AOPS Precalculus Solutions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

AOPS Precalculus Solutions

Uploaded by

Copyright:

Available Formats

PHILLIPS ACADEMY

# OLIVER • WENDELL • HOLMES

Art of Problem Solving

Published by: AoPS Incorporated

Cover image designed by Vanessa Rusczyk using KaleidoTile software.

Printed in the United States of America.

First Printing 2009.

Introduction to Trigonometric Functions 14

Parameterization and Trigonometric Coordinate Systems 86

6 Basics of Complex Numbers 104

7 Trigonometry and Complex Numbers 123

Geometry of Complex Numbers 151

Vectors in Two Dimensions 177

1 Matrices in Two Dimensions 187

1 1 Vectors and Matrices in Three Dimensions, Part 1 207

Exercises for Section 1.1

(a) /(2) = 22 - 2 - 6 = 1-4 |.

(b) Since f(x) = x2 - x - 6 = (x - 3)(x + 2), we have /(x) = 0 when x = 3 and x - -2

(c) f(x - 1) = (x - l)2 - (x - 1) - 6 = x2 - 2x + 1 - x + 1 - 6 = 3x - 4

range of / is [0, +oo) .

As with / • g, the function

1.1.4 The function

On the other hand, the function

(a) We have T( 2,3, -5) = 3 • 23 - (-5) = 3-8 + 5 = [~29

Exercises for Section 1.2

(c) Note that /(x) = x2 - 5x + 6 factors as (x - 2)(x - 3). If (x - 2)(x - 3) = 0, then y

Exercises for Section 1.3

domain of /. For x < 1, we therefore have f(g(x)) = /(Vl - x) = 1 - x - 2 Vl - x, where x < 1

f\x) = f(x) = ax,

Exercises for Section 1.4

^=3 * y - 1 = 3(y - 2) = 3y - 6 => y

(a) Let x = /(y) = 2y - 7. Then y = (x + 7)/2, so / x(x) = me + 7)/2

x(cy + d) = ay + b => cxy + dx = ay + b => cxy - ay = b - dx => y(cx - a) = b — dx

The results for all five parts are shown below.

cannot represent a function. / :4: _

1.31 If /(x) = -x2 + bx + c and g(x) = dx + e, then

1.34 If y =/(x) = x/(l - x), then

y(l-x) = x =» y - xy - x =» xy + x = y => x(y + 1) = y => x= y

We compare this to each of the given possibilities:

Therefore, we have x = -/(-y)

If /(/(x)) = x for all x except -3/2, then

This simplifies to x as long as ab + 4 =£ 0. If ab + 4 = 0, then a = -4/b, and

2x + a _ 2x - 4/b _ 2bx - 4 _ 2(fcx - 2) _ 2

which is a constant, so / does not have an inverse if ab = -4.

Therefore, the domain of /(1/x) is (—oo, -1) u (1, +oo)

1.41 Let t = x2 + 1, so x2 = t - 1 and x4 = (t - l)2. Hence,

a ■ (b - dx)/(cx - a) + b _ a(b - dx) + b(cx - a) (be - ad)x

Now we compute /(/(x)):

ax + b\ a - (ax + b)/(cx + d) + b a(ax + b) + b(cx + d)

(a1 + bc)x + (ab + bd)

Also, /(97) = 97, so we have

Subtracting the first equation from the second, we get

Exercises for Section 2.1

cos 60° OS _ 1/2 V3

(e) The terminal point of 180° is (-1,0), so cos 180° = -1

(a) The terminal point of -90° is (0, -1), so cos(-90°) - 0

sec(-1200°) = sec 240° = -2

Thus, cos(-68°), cot 206°, and esc 90.5° are positive.

Exercises for Section 2.2